# gpt-oss

OpenAI open-weight reasoning language models for local and infrastructure deployment.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Probing the Trajectories of Reasoning Traces in Large Language Models

DevFeed: [Probing the Trajectories of Reasoning Traces in Large Language Models](<https://devfeed.tech/articles/probing-the-trajectories-of-reasoning-traces-in-large-language-models-138736.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2601.23163>)

Author: Marthe Ballon, Brecht F. Verbeken, Vincent Ginis, Andres Algaba

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Chain-of-thought](<https://devfeed.tech/topics/chain-of-thought.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [qwen](<https://devfeed.tech/topics/qwen.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [deepseek-r1](<https://devfeed.tech/tags/deepseek-r1.md>), [gemma](<https://devfeed.tech/tags/gemma.md>), [gemma-4](<https://devfeed.tech/tags/gemma-4.md>), [gemma-4-31b](<https://devfeed.tech/tags/gemma-4-31b.md>), [gpt-oss](<https://devfeed.tech/tags/gpt-oss.md>), [gpt-oss-20b](<https://devfeed.tech/tags/gpt-oss-20b.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [qwen3](<https://devfeed.tech/tags/qwen3.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [reasoning-model](<https://devfeed.tech/tags/reasoning-model.md>), [traces](<https://devfeed.tech/tags/traces.md>)

### AI overview

The proposed trajectory-probing protocol measures how a model's answer distribution changes as it receives progressively longer portions of its own reasoning trace: generate a full trace, truncate it at token percentiles, then feed each partial trace back and measure the induced answers. Across seven open-source models and four benchmarks, accuracy and decision commitment generally rise with the amount of reasoning provided. Length, form, and token-identity controls indicate that the gains come from instance-specific semantic content, not simply longer context or generic reasoning style. In cross-model probes, even stronger models often reproduce errors in a source model's partial trace, while allowing additional reasoning substantially improves recovery. The abstract presents the method as model-agnostic and capable of exposing failure modes that aggregate accuracy can miss.

### Source excerpt

arXiv:2601.23163v2 Announce Type: replace Abstract: Large language models (LLMs) increasingly solve difficult problems by producing reasoning traces before emitting a final response. However, it remains unclear how accuracy and decision commitment evolve along a reasoning trajectory, and whether intermediate trace segments provide answer-relevant information beyond generic length or stylistic effects. Here, we propose a trajectory-probing protocol for evaluating how a reasoning model's predicted answer evolves along a partial reasoning trace. The protocol (1) generates a model's full reasoning trace, (2) truncates it at fixed token-percentiles, and (3) injects each partial trace back into the model, measuring the model's induced answer distribution. We apply the protocol to seven open-source reasoning models (Qwen3-4B/8B/14B, gpt-oss-20b/120b, DeepSeek-R1-Distill-Llama-8B and Gemma-4-31B-it) across four benchmarks (GPQA Diamond, MMLU-Pro, Omni-MATH-2 and CompMath). The protocol reveals that accuracy and decision commitment generally increase as the percentage of provided reasoning tokens grows, and length, form, and token-identity controls confirm that gains stem from instance-specific semantic content rather than context length or generic reasoning style effects. Furthermore, cross-model probing shows that even stronger models frequently repeat the source model's errors when given its partial reasoning trace, while allowing further reasoning substantially improves recovery. Together, these results provide a reproducible, model-agnostic measurement of reasoning trace dynamics that generalizes across answer formats and can reveal model-specific failure modes invisible to aggregate accuracy.

## The Logic Chain Breaking Point: A Controlled-Variable Benchmark for Reasoning Degradation in Open-Weight Language Models

DevFeed: [The Logic Chain Breaking Point: A Controlled-Variable Benchmark for Reasoning Degradation in Open-Weight Language Models](<https://devfeed.tech/articles/the-logic-chain-breaking-point-a-controlled-variable-benchmark-for-reasoning-degradation-in-open-weight-language-models-137573.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20447>)

Author: Akif Islam, Mohd. Ruhul Ameen, Md. Ekramul Hamid

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [logic](<https://devfeed.tech/tags/logic.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>)

### AI overview

In this benchmark, detecting inconsistent premises causes the strongest reasoning degradation: instruction-tuned models, including 24B-27B baselines, score essentially zero, while reasoning-tuned models recover only partially and remain below reliable accuracy. By comparison, most models stay near saturation on pure modus-ponens chains up to length 128, and distractors cause model-specific declines mainly at high levels. The study tests 16 open-weight models across chain scaling, distractor resilience, and contradiction placement, using greedy decoding and known-by-construction answers. Trials limited by context or output length are excluded from valid accuracy. The reported degradation on inconsistent premises is specific to the benchmark's abstention-based scoring.

### Source excerpt

arXiv:2611.20447v1 Announce Type: new Abstract: We introduce The Logic Chain Breaking Point, a controlled-variable benchmark for deductive reasoning in open-weight language models. It spans three phases: (i) pure modus-ponens chain scaling, (ii) distractor resilience with a fixed core chain, and (iii) contradiction detection under primacy versus recency placement. Independent variables follow a powers-of-two sequence $v \in \{1,2,4,8,16,32,64,128\}$, with ground truth known by construction. We evaluate sixteen open-weight models across instruction-tuned and reasoning-tuned panels, including a small-model suite and a larger local-model extension. All models are evaluated with greedy decoding under model-appropriate context and output budgets, while context- and length-limited trials are quarantined from the valid accuracy denominator. Our results reveal a sharply non-monotonic difficulty surface: pure deductive chains up to $L=128$ remain near saturation for most models, and distractor noise causes only model-specific degradation at high $D$, while the inconsistent premise condition has the lowest accuracy. Instruction-tuned models, including larger 24B--27B baselines, score essentially zero on inconsistent premise sets, while reasoning-tuned models recover only partially and remain below reliable accuracy. Within this benchmark, inconsistent premise sets are associated with the strongest degradation under the benchmark's abstention-based scoring rule.

## Local LLMs Handle Textbook RAG but Fail on New Research Papers

DevFeed: [Local LLMs Handle Textbook RAG but Fail on New Research Papers](<https://devfeed.tech/articles/local-llms-handle-textbook-rag-but-fail-on-new-research-papers-135976.md>)

Original publisher: [Read original article](<https://towardsdatascience.com/local-llms-handle-textbook-rag-but-fail-on-new-research-papers/>)

Author: Waleed Esmail

Published: 2026-11-16T12:00:00Z

Content type: article

Language: en

Sources: [Towards Data Science](<https://devfeed.tech/sources/towards-data-science.md>)

Topics: [Local AI](<https://devfeed.tech/topics/local-ai.md>), [generalization in machine learning](<https://devfeed.tech/topics/generalization-in-machine-learning.md>), [Hallucination detection](<https://devfeed.tech/topics/hallucination-detection.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [llms](<https://devfeed.tech/tags/llms.md>), [local](<https://devfeed.tech/tags/local.md>), [local-llm-inference](<https://devfeed.tech/tags/local-llm-inference.md>), [local-llms](<https://devfeed.tech/tags/local-llms.md>), [papers](<https://devfeed.tech/tags/papers.md>), [rag](<https://devfeed.tech/tags/rag.md>), [rag-retrieval-augmented-generation](<https://devfeed.tech/tags/rag-retrieval-augmented-generation.md>)

### AI overview

In this evaluation, five setups scored 12-20 of 20 on textbook questions but only 0-9 of 15 on a paper published that year, despite identical retrieval and prompts. The author concludes that small local models handle familiar material better than new research, where they must rely on supplied passages. Grading answers alongside retrieved passages showed many errors came from missing evidence: figure descriptions raised figure-question results from 1/10 to 9/10, chapter summary trees raised overview results from 1/8 to 8/8, and neighboring chunks improved complete retrieval from 16/20 to 19/20. Quantization did not explain the gap. Qwen3.8-27B was close to Opus in the reported comparison at lower cost, but took about three minutes per answer. The author recommends writing gold questions first, separating familiar from new material, inspecting passages when grading, improving retrieval before changing models, and routing figures, web reading, and derivations to stronger models. Results are limited by a 35-question evaluation and grader drift.

### Source excerpt

Five setups, one blind grading pack: textbook scores ran 12 to 20 of 20, but a paper published this year dropped them to 0 to 9 of 15. The post Local LLMs Handle Textbook RAG but Fail on New Research Papers appeared first on Towards Data Science.

## Do Large Language Models Show Racial and Sex Bias in Pain Management? A Counterfactual Audit of GPT, Claude, Gemini, Llama, DeepSeek and Qwen

DevFeed: [Do Large Language Models Show Racial and Sex Bias in Pain Management? A Counterfactual Audit of GPT, Claude, Gemini, Llama, DeepSeek and Qwen](<https://devfeed.tech/articles/do-large-language-models-show-racial-and-sex-bias-in-pain-management-a-counterfactual-audit-of-gpt-claude-gemini-llama-deepseek-and-qwen-132069.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14935>)

Author: Richard J. Young, Alice M. Matthews

Published: 2026-11-13T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [llama](<https://devfeed.tech/topics/llama.md>), [qwen](<https://devfeed.tech/topics/qwen.md>)

Tags: [17](<https://devfeed.tech/tags/17.md>), [2](<https://devfeed.tech/tags/2.md>), [about](<https://devfeed.tech/tags/about.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [audit](<https://devfeed.tech/tags/audit.md>), [bias](<https://devfeed.tech/tags/bias.md>), [claude](<https://devfeed.tech/tags/claude.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cy](<https://devfeed.tech/tags/cs-cy.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [gpt](<https://devfeed.tech/tags/gpt.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large](<https://devfeed.tech/tags/large.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [llama](<https://devfeed.tech/tags/llama.md>), [management](<https://devfeed.tech/tags/management.md>), [models](<https://devfeed.tech/tags/models.md>), [pain](<https://devfeed.tech/tags/pain.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [show](<https://devfeed.tech/tags/show.md>)

### AI overview

In exploratory, non-preregistered tests using a single JSON-constrained prompt, the audit found no race- or sex-related differences that survived correction across four open-weight models or 24 configurations of 17 models. Pain-severity differences were within one point on a 0-10 scale, and within a quarter point for 237 of 238 Q-Pain comparisons. Opioid decisions were less precisely bounded: Study 1 could not rule out differences of about 20 percentage points, so no detected difference is not proof of none. A few corrected within-model differences favored Black patients. Variation between models and their settings was much larger than variation between patient identities; the White male reference received the high opioid dose in 6% to 93% of responses. Serving software, provider and prompt format also varied, limiting attribution. The authors recommend multiple names per group, seed-paired comparisons, measuring response noise and auditing each deployed configuration.

### Source excerpt

arXiv:2611.14935v1 Announce Type: cross Abstract: Racial and ethnic minority patients receive less pain treatment than non-Hispanic White patients, and large language models (LLMs) used as medical AI could reproduce that bias. Prior audits tested a few proprietary models, stated race explicitly and rarely measured the models' own response instability. In exploratory, non-preregistered analyses, Study 1 had four open-weight LLMs assess six acute-pain vignettes in every combination of four races and two sexes, with identity stated or cued only by a name (two names per group), comparing each identity with a White male reference on the same case and sampling seed. In Study 2, 24 configurations of 17 models, including OpenAI's GPT-4.1 mini and gpt-oss, Anthropic's Claude Haiku 4.5, Google's Gemini 2.5 Flash, Meta's Llama, DeepSeek and Qwen, with reasoning on and off and at two precisions, answered the same vignettes and the 50 clinician-written Q-Pain cases. In this single-prompt, JSON-constrained task, no difference survived correction in Study 1, none of 1,386 comparisons survived the false discovery rate in Study 2, and every pain-severity difference was equivalent to zero within one point on a 0-10 scale (within a quarter point for 237 of 238 Q-Pain comparisons). Opioid decisions were bounded less precisely (Study 1 could not exclude differences of about 20 percentage points), so they show no detected difference, not proof of none. The few differences surviving correction within a model favoured Black patients. Models differed far more from one another and across their own settings than between patients (the reference patient received the high opioid dose in 6% to 93% of responses), although serving software, provider and prompt format also differed. Fairness audits of clinical AI should use several names per group, pair comparisons on the seed, report the noise floor and test each deployed configuration.

## How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning

DevFeed: [How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning](<https://devfeed.tech/articles/how-much-does-the-hint-matter-scope-conditions-for-hindsight-self-distillation-in-financial-reasoning-103082.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.36082>)

Author: Johnny Yu, Cyril Gorlla

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Finance](<https://devfeed.tech/topics/finance.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [accounting](<https://devfeed.tech/tags/accounting.md>), [audit](<https://devfeed.tech/tags/audit.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data-residency](<https://devfeed.tech/tags/data-residency.md>), [distillation](<https://devfeed.tech/tags/distillation.md>), [error](<https://devfeed.tech/tags/error.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [financial](<https://devfeed.tech/tags/financial.md>), [financial-institutions](<https://devfeed.tech/tags/financial-institutions.md>), [gpt-oss](<https://devfeed.tech/tags/gpt-oss.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [hint](<https://devfeed.tech/tags/hint.md>), [hints](<https://devfeed.tech/tags/hints.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [regression](<https://devfeed.tech/tags/regression.md>), [student](<https://devfeed.tech/tags/student.md>)

### AI overview

Fin-SD improves multi-step financial reasoning by using a model's own failed attempt to focus distillation on the point just before its first error. The student critiques that step; a frozen copy conditions on the preceding reasoning and a hint, then trains the student over a short rollout with reverse KL, without an external model contributing gradients. On GPT-OSS-20B, generic feedback produced the same 3.03 percentage-point gain on FinanceReasoning Hard as the critique, while hints from GPT-OSS-120B offered no material benefit over self-written hints. Fin-SD matched full-reference-solution self-distillation in accuracy with 9.5% fewer tokens. The reported MMLU-Pro and 14-metric FinTrust checks found no material regression.

### Source excerpt

arXiv:2610.36082v1 Announce Type: cross Abstract: Financial institutions operating under data-residency constraints are restricted to open-weight models small enough to locally host, which trail frontier systems on multi-step quantitative reasoning. On-policy distillation narrows such gaps by supervising a student on its own rollouts, but traditional OPD requires a strong external teacher, and on-policy self-distillation (OPSD) conditions on a full reference solution. We ask how much of that signal is necessary. We present Fin-SD, a localized hindsight self-distillation recipe for multi-step reasoning. Given a problem the student fails, the student itself identifies the first incorrect reasoning step and writes a short critique of that step. We truncate the student's trajectory immediately before the error (aiming to obtain a correct prefix), condition a frozen copy of the student on the same prefix plus the hint, and distill that hint-conditioned teacher into the student over a short rollout under reverse KL. No external model contributes a gradient. We apply Fin-SD to GPT-OSS-20B on a custom multi-step financial reasoning benchmark that requires deriving formula inputs from realistic scenarios via accounting conventions. We find: (i) replacing the critique with a generic-feedback control matches Fin-SD's entire 3.03 pp accuracy gain over the base model on the FinanceReasoning Hard subset, suggesting the gain comes from where supervision is applied, not hint content; (ii) the model writing the hint does not need to be stronger than the model receiving it: hints from same-family GPT-OSS-120B, at roughly 6x the parameters, yield no materially greater benefits than self-written hints; and (iii) Fin-SD matches on-policy self-distillation from full reference solutions on accuracy while generating 9.5% fewer tokens. With Fin-SD, general reasoning on MMLU-Pro and a fourteen-metric compliance audit on FinTrust show no material regression.

## Training an MoE LLM in MXFP4: How We Closed the Gap to BF16

DevFeed: [Training an MoE LLM in MXFP4: How We Closed the Gap to BF16](<https://devfeed.tech/articles/training-an-moe-llm-in-mxfp4-how-we-closed-the-gap-to-bf16-99974.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.32229>)

Author: Han Wang, Zhitao Wang, Natalia Frumkin, Xuanwu Yin, Alireza Khodamoradi, Dong Li, Emad Barsoum

Published: 2026-10-26T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [fp4](<https://devfeed.tech/tags/fp4.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llm](<https://devfeed.tech/tags/llm.md>), [mixture-of-experts](<https://devfeed.tech/tags/mixture-of-experts.md>), [mixture-of-experts-moe](<https://devfeed.tech/tags/mixture-of-experts-moe.md>), [mlperf](<https://devfeed.tech/tags/mlperf.md>), [moe](<https://devfeed.tech/tags/moe.md>), [mxfp4](<https://devfeed.tech/tags/mxfp4.md>), [results](<https://devfeed.tech/tags/results.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

ALTO's MXFP4 training recipe brings GPT-OSS-20B close to its BF16 validation result: on C4 at global batch size 16, MXFP4 with randomized Hadamard transforms, stochastic rounding, and weight de-oscillation reached loss 3.3350 at 16,128 steps, versus 3.3283 for BF16. De-oscillation closed about half the gap versus MXFP4 without it (loss 3.3418) and reduced FP4's extra steps to reach the MLPerf loss target of 3.34 from 20% to 5% over BF16. The recipe combines hybrid block quantization and other established techniques with a simplified de-oscillation method. Differential gradient estimation, outlier clipping, and macro-block scaling improved operator-level signal-to-noise ratio but hurt end-to-end results. Experiments used simulated MXFP4 kernels on AMD MI300; wall-clock performance was not optimized.

### Source excerpt

arXiv:2610.32229v1 Announce Type: new Abstract: Training large language models (LLMs) in sub-8-bit arithmetic promises substantial gains in memory bandwidth and compute throughput, yet FP4 formats introduce quantization error severe enough to destabilize optimization if left unaddressed. We present ALTO, an open-source training recipe for GPT-OSS-20B, a 20-billion-parameter mixture-of-experts (MoE) model, on the MLPerf Small MoE Training benchmark under the MXFP4 microscaling format. The recipe combines established techniques from Abecassis, Felix et al. [arXiv:2509.25149] (hybrid 1D/2D block quantization, randomized Hadamard transforms, and stochastic rounding) with a simplified weight de-oscillation scheme adapted from TetraJet-v2 [arXiv:2510.27527]. On C4 validation at global batch size 16, MXFP4 + RHT + SR + de-oscillation reaches a validation loss of 3.3350 at 16,128 steps, only 0.007 above the BF16 baseline (3.3283), closing roughly 50% of the gap left by MXFP4 without de-oscillation (3.3418). Measured by steps to reach the MLPerf quality target (validation loss 3.34), de-oscillation cuts the FP4 convergence overhead relative to BF16 from +20.0% to +5.0%. We also report negative end-to-end results from differential gradient estimation, outlier clipping, and macro-block scaling, despite each showing operator-level SNR gains. All experiments use simulated MXFP4 kernels on AMD MI300 hardware and wall-clock speed has not been optimized.

## You.com Launches Inference Router: 12 Open-Weight Models With 99.9% Uptime SLA

DevFeed: [You.com Launches Inference Router: 12 Open-Weight Models With 99.9% Uptime SLA](<https://devfeed.tech/articles/you-com-launches-inference-router-12-open-weight-models-with-99-9-uptime-sla-97209.md>)

Original publisher: [Read original article](<https://www.marktechpost.com/2026/10/22/you-com-inference-router-open-weight-models-uptime-sla/>)

Author: Asif Razzaq

Published: 2026-10-23T05:28:21Z

Content type: news

Language: en

Sources: [MarkTechPost](<https://devfeed.tech/sources/marktechpost.md>)

Topics: [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-shorts](<https://devfeed.tech/tags/ai-shorts.md>), [api](<https://devfeed.tech/tags/api.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [data](<https://devfeed.tech/tags/data.md>), [data-retention](<https://devfeed.tech/tags/data-retention.md>), [editors-pick](<https://devfeed.tech/tags/editors-pick.md>), [enterprise-ai](<https://devfeed.tech/tags/enterprise-ai.md>), [first](<https://devfeed.tech/tags/first.md>), [for-devs](<https://devfeed.tech/tags/for-devs.md>), [hosting](<https://devfeed.tech/tags/hosting.md>), [inference](<https://devfeed.tech/tags/inference.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [launches](<https://devfeed.tech/tags/launches.md>), [models](<https://devfeed.tech/tags/models.md>), [new-releases](<https://devfeed.tech/tags/new-releases.md>), [open](<https://devfeed.tech/tags/open.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [openai](<https://devfeed.tech/tags/openai.md>), [platform](<https://devfeed.tech/tags/platform.md>), [post](<https://devfeed.tech/tags/post.md>), [retention](<https://devfeed.tech/tags/retention.md>), [router](<https://devfeed.tech/tags/router.md>), [search](<https://devfeed.tech/tags/search.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [staff](<https://devfeed.tech/tags/staff.md>), [tech-news](<https://devfeed.tech/tags/tech-news.md>), [technology](<https://devfeed.tech/tags/technology.md>), [uptime](<https://devfeed.tech/tags/uptime.md>), [weight](<https://devfeed.tech/tags/weight.md>), [with](<https://devfeed.tech/tags/with.md>), [you-com](<https://devfeed.tech/tags/you-com.md>)

### AI overview

You.com's Inference Router provides one OpenAI- and Anthropic-compatible API for 12 open-weight models, with pay-as-you-go pricing, no platform fee, a 99.9% minimum uptime SLA, and US-hosted failover from You.com Inference Direct to named partners. Its zero data retention policy covers prompts and outputs for both model calls and web search; it retains operational metadata. Developers can enable server-side search and extraction, with citations in results. The article reports self-run LiveNewsBench comparisons suggesting lower costs for some models, but labels those results self-reported. It says token prices are generally near market, so the stated value centers on hosting, retention, uptime and routing guarantees.

### Source excerpt

You.com launched Inference Router, an OpenAI-compatible API for 12 open-weight models with a 99.9% uptime SLA, zero data retention, 100% US hosting, server-side web search and no platform fee. The post You.com Launches Inference Router: 12 Open-Weight Models With 99.9% Uptime SLA appeared first on MarkTechPost.

## Fine-tuning GPT-OSS-20B for multilingual reasoning on Amazon SageMaker HyperPod

DevFeed: [Fine-tuning GPT-OSS-20B for multilingual reasoning on Amazon SageMaker HyperPod](<https://devfeed.tech/articles/fine-tuning-gpt-oss-20b-for-multilingual-reasoning-on-amazon-sagemaker-hyperpod-91293.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/machine-learning/fine-tuning-gpt-oss-20b-for-multilingual-reasoning-on-amazon-sagemaker-hyperpod/>)

Author: Mijanur Palash

Published: 2026-10-20T17:48:27Z

Content type: tutorial

Language: en

Sources: [Artificial Intelligence](<https://devfeed.tech/sources/artificial-intelligence.md>)

Topics: [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Amazon SageMaker HyperPod](<https://devfeed.tech/topics/amazon-sagemaker-hyperpod.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [lora](<https://devfeed.tech/topics/lora.md>), [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Enhancing AI model performance](<https://devfeed.tech/topics/enhancing-ai-model-performance.md>), [trl](<https://devfeed.tech/topics/trl.md>), [peft](<https://devfeed.tech/topics/peft.md>), [Amazon SageMaker](<https://devfeed.tech/topics/amazon-sagemaker.md>), [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>)

Tags: [adapter](<https://devfeed.tech/tags/adapter.md>), [adapters](<https://devfeed.tech/tags/adapters.md>), [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [amazon](<https://devfeed.tech/tags/amazon.md>), [amazon-eks](<https://devfeed.tech/tags/amazon-eks.md>), [amazon-elastic-container-registry](<https://devfeed.tech/tags/amazon-elastic-container-registry.md>), [amazon-elastic-kubernetes-service](<https://devfeed.tech/tags/amazon-elastic-kubernetes-service.md>), [amazon-elastic-kubernetes-service-amazon-eks](<https://devfeed.tech/tags/amazon-elastic-kubernetes-service-amazon-eks.md>), [amazon-fsx](<https://devfeed.tech/tags/amazon-fsx.md>), [amazon-fsx-for-lustre](<https://devfeed.tech/tags/amazon-fsx-for-lustre.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [amazon-sagemaker](<https://devfeed.tech/tags/amazon-sagemaker.md>), [amazon-sagemaker-hyperpod](<https://devfeed.tech/tags/amazon-sagemaker-hyperpod.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [gpt-oss](<https://devfeed.tech/tags/gpt-oss.md>), [grpo](<https://devfeed.tech/tags/grpo.md>), [headroom](<https://devfeed.tech/tags/headroom.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [sft](<https://devfeed.tech/tags/sft.md>), [supervised-fine-tuning](<https://devfeed.tech/tags/supervised-fine-tuning.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>)

### AI overview

Fine-tuning GPT-OSS-20B with LoRA on Amazon SageMaker HyperPod raised reasoning in the requested language from 22% to 100% on a 50-case, five-language evaluation. SFT taught the missing behavior; final-answer language accuracy reached 62%. GRPO on the SFT checkpoint raised that metric to 70% while maintaining 100% reasoning-language accuracy. A Qwen2.5-7B control showed GRPO added 6 points on multilingual reasoning where there was headroom, but no gain on GSM8K, already at 90.5%. The results support using SFT for behaviors the model does not yet exhibit and adding GRPO to refine behaviors it sometimes produces; when sampled groups are already mostly correct, GRPO may have no useful learning signal. The article describes HyperPod training infrastructure and configuration, including TRL for GPT-OSS-20B and veRL with Ray and vLLM for the Qwen control.

### Source excerpt

Learn how we fine-tuned GPT-OSS-20B for multilingual reasoning on Amazon SageMaker HyperPod, and how to decide between supervised fine-tuning (SFT) and reinforcement learning with GRPO. SFT took reasoning-language accuracy from 22% to 100%, and GRPO added measurable gains where headroom existed.

## How Much GPU Memory Do You Really Need to Run an Open LLM Locally?

DevFeed: [How Much GPU Memory Do You Really Need to Run an Open LLM Locally?](<https://devfeed.tech/articles/how-much-gpu-memory-do-you-really-need-to-run-an-open-llm-locally-89725.md>)

Original publisher: [Read original article](<https://hackernoon.com/how-much-gpu-memory-do-you-really-need-to-run-an-open-llm-locally?source=rss>)

Author: Contexte tech

Published: 2026-10-18T13:12:25Z

Content type: tutorial

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [DGX Spark](<https://devfeed.tech/topics/dgx-spark.md>)

Tags: [gpu-memory](<https://devfeed.tech/tags/gpu-memory.md>), [gpu-memory-requirements](<https://devfeed.tech/tags/gpu-memory-requirements.md>), [kv-cache](<https://devfeed.tech/tags/kv-cache.md>), [llm-hardware-sizing](<https://devfeed.tech/tags/llm-hardware-sizing.md>), [llm-quantization](<https://devfeed.tech/tags/llm-quantization.md>), [local-llm](<https://devfeed.tech/tags/local-llm.md>), [local-llm-hardware](<https://devfeed.tech/tags/local-llm-hardware.md>), [memory-bandwidth](<https://devfeed.tech/tags/memory-bandwidth.md>), [mixture-of-experts-models](<https://devfeed.tech/tags/mixture-of-experts-models.md>), [self-hosted-ai](<https://devfeed.tech/tags/self-hosted-ai.md>), [vram-estimation](<https://devfeed.tech/tags/vram-estimation.md>)

### AI overview

The tutorial explains how to estimate local LLM hardware requirements from model weights, KV cache size, memory bandwidth and request volume. It provides formulas, model and hardware comparisons, and Python calculations. It distinguishes theoretical throughput ceilings from measured performance and recommends explicit cache calculations for long contexts or concurrent users, followed by workload-specific benchmarking.

### Source excerpt

How to size hardware for a local LLM: weights, KV cache and memory bandwidth, checked against official specs for 12 open models and 9 GPUs and Macs.

## Vertex AI Shutdowns This Week: Gemini 2.5 on October 20, 16 Open Models on October 21

DevFeed: [Vertex AI Shutdowns This Week: Gemini 2.5 on October 20, 16 Open Models on October 21](<https://devfeed.tech/articles/vertex-ai-shutdowns-this-week-gemini-2-5-on-october-20-16-open-models-on-october-21-89714.md>)

Original publisher: [Read original article](<https://www.developersdigest.tech/blog/vertex-ai-shutdowns-october-2026>)

Author: Developers Digest

Published: 2026-10-18T00:00:00Z

Content type: news

Language: en

Sources: [Developers Digest](<https://devfeed.tech/sources/developers-digest.md>)

Topics: [Google AI](<https://devfeed.tech/topics/google-ai.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [deprecations](<https://devfeed.tech/tags/deprecations.md>), [flash-lite](<https://devfeed.tech/tags/flash-lite.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [google](<https://devfeed.tech/tags/google.md>), [ids](<https://devfeed.tech/tags/ids.md>), [migration](<https://devfeed.tech/tags/migration.md>), [prices](<https://devfeed.tech/tags/prices.md>), [vertex-ai](<https://devfeed.tech/tags/vertex-ai.md>)

### AI overview

Google Cloud is retiring Gemini 2.5 Pro, Flash, and Flash-Lite on Vertex AI on October 20, 2026, followed by 16 open MaaS models the next day. The article covers replacement model IDs, prices, and code changes.

### Source excerpt

Google Cloud retires Gemini 2.5 Pro, Flash and Flash-Lite on Vertex AI on October 20, 2026, and 16 open MaaS models the next day. Replacement ids, prices, and the code change.

## Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context

DevFeed: [Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context](<https://devfeed.tech/articles/meta-releases-muse-spark-open-816b-open-weight-moe-with-1m-context-88270.md>)

Original publisher: [Read original article](<https://www.marktechpost.com/2026/10/15/meta-muse-spark-open-816b-open-weight-model/>)

Author: Asif Razzaq

Published: 2026-10-16T02:26:04Z

Content type: news

Language: en

Sources: [MarkTechPost](<https://devfeed.tech/sources/marktechpost.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [TensorRT-LLM](<https://devfeed.tech/topics/tensorrt-llm.md>), [vercel ai sdk](<https://devfeed.tech/topics/vercel-ai-sdk.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-shorts](<https://devfeed.tech/tags/ai-shorts.md>), [apple](<https://devfeed.tech/tags/apple.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [context-window](<https://devfeed.tech/tags/context-window.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [editors-pick](<https://devfeed.tech/tags/editors-pick.md>), [for-devs](<https://devfeed.tech/tags/for-devs.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [moe](<https://devfeed.tech/tags/moe.md>), [muse-spark](<https://devfeed.tech/tags/muse-spark.md>), [new-releases](<https://devfeed.tech/tags/new-releases.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [staff](<https://devfeed.tech/tags/staff.md>), [tech-news](<https://devfeed.tech/tags/tech-news.md>), [technology](<https://devfeed.tech/tags/technology.md>)

### AI overview

Meta released Muse Spark Open, an 816B-parameter open-weight multimodal model with 42B active parameters and a 1M-token context window. The article reports strong results on tool-use benchmarks, weaker coding performance than several competing models, and deployment options ranging from multi-GPU systems to two M5 Ultra Mac Studios. It also summarizes available inference and fine-tuning engines and key license conditions.

### Source excerpt

Meta released Muse Spark Open, an 816B-parameter open-weight MoE model with 42B active parameters and a 1M-token context. It leads open models on MCP-Atlas and runs on 2 M5 Ultra Mac Studios. The post Meta Releases Muse Spark Open: 816B Open-Weight MoE With 1M Context appeared first on MarkTechPost.

## Meta releases Muse Spark Open, a new 'open-weight' AI model

DevFeed: [Meta releases Muse Spark Open, a new 'open-weight' AI model](<https://devfeed.tech/articles/meta-releases-muse-spark-open-a-new-open-weight-ai-model-88604.md>)

Original publisher: [Read original article](<https://www.theguardian.com/technology/2026/oct/16/meta-releases-muse-spark-open-a-new-open-weight-ai-model>)

Author: Agence France-Presse

Published: 2026-10-16T00:34:28Z

Content type: news

Language: en

Sources: [The Guardian | Technology](<https://devfeed.tech/sources/the-guardian-technology.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-artificial-intelligence](<https://devfeed.tech/tags/ai-artificial-intelligence.md>), [ai-artificial-intelligence-technology-meta-china-us-news-computing](<https://devfeed.tech/tags/ai-artificial-intelligence-technology-meta-china-us-news-computing.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [alexandr-wang](<https://devfeed.tech/tags/alexandr-wang.md>), [china](<https://devfeed.tech/tags/china.md>), [coding](<https://devfeed.tech/tags/coding.md>), [computing](<https://devfeed.tech/tags/computing.md>), [meta](<https://devfeed.tech/tags/meta.md>), [moonshot-ai](<https://devfeed.tech/tags/moonshot-ai.md>), [muse](<https://devfeed.tech/tags/muse.md>), [muse-spark](<https://devfeed.tech/tags/muse-spark.md>), [muse-spark-open](<https://devfeed.tech/tags/muse-spark-open.md>), [open-weight-ai](<https://devfeed.tech/tags/open-weight-ai.md>), [releases](<https://devfeed.tech/tags/releases.md>), [software](<https://devfeed.tech/tags/software.md>), [technology](<https://devfeed.tech/tags/technology.md>), [us-news](<https://devfeed.tech/tags/us-news.md>)

### AI overview

Meta has launched Muse Spark Open, an open-weight AI model it says competes with leading Chinese models and is the most capable US open-weight model to date. Meta says it is built for coding, agentic work, complex software engineering and multi-step reasoning. The model is optimized for Apple M5 Ultra chips and can run across two Mac Studios, though it requires significant computing power.

### Source excerpt

Meta says Muse excels at 'multi-step reasoning' and 'complex software engineering' as it races to catch up to Chinese rivals Meta on Thursday launched a new open-weight AI model which it claims rivals, and in some cases outperforms, similar products from leading Chinese developers. Muse Spark Open "is the most capable US open-weight model to date," Meta's chief of AI Alexandr Wang wrote on X. Continue reading...

## Reflection's Beam is freely available on Ollama for a limited time

DevFeed: [Reflection's Beam is freely available on Ollama for a limited time](<https://devfeed.tech/articles/reflection-s-beam-is-freely-available-on-ollama-for-a-limited-time-88804.md>)

Original publisher: [Read original article](<https://ollama.com/blog/reflection-beam>)

Published: 2026-10-16T00:00:00Z

Content type: release

Language: en

Sources: [Ollama](<https://devfeed.tech/sources/ollama.md>)

Topics: [Ollama](<https://devfeed.tech/topics/ollama.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Loop Engineering](<https://devfeed.tech/topics/loop-engineering.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [agents](<https://devfeed.tech/tags/agents.md>), [apache-2-0](<https://devfeed.tech/tags/apache-2-0.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [model](<https://devfeed.tech/tags/model.md>), [ollama](<https://devfeed.tech/tags/ollama.md>)

### AI overview

Reflection's first open-weight model, Beam, is available through Ollama Cloud for free for a limited time. It is designed for software engineering, reasoning, and long-running agentic workflows, and Ollama says its endpoint uses zero data retention.

### Source excerpt

Reflection's first open-weight model is now available on Ollama, and free for a limited time only. Beam is designed for software engineering, reasoning, and long-running agentic tasks. Access Beam through Ollama Cloud.

## Meta releases 816B-parameter Muse Spark Open for coding and agent work

DevFeed: [Meta releases 816B-parameter Muse Spark Open for coding and agent work](<https://devfeed.tech/articles/meta-releases-816b-parameter-muse-spark-open-for-coding-and-agent-work-88287.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/10/15/meta-releases-816b-parameter-muse-spark-open-for-coding-and-agent-work/>)

Author: Duncan Riley

Published: 2026-10-15T22:20:54Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [agent orchestration](<https://devfeed.tech/topics/agent-orchestration.md>), [agent to agent communication](<https://devfeed.tech/topics/agent-to-agent-communication.md>), [cursor](<https://devfeed.tech/topics/cursor.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [amazon-web-services](<https://devfeed.tech/tags/amazon-web-services.md>), [api](<https://devfeed.tech/tags/api.md>), [apple](<https://devfeed.tech/tags/apple.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [coding](<https://devfeed.tech/tags/coding.md>), [davinci-resolve](<https://devfeed.tech/tags/davinci-resolve.md>), [fireworks-ai](<https://devfeed.tech/tags/fireworks-ai.md>), [glm-5-3](<https://devfeed.tech/tags/glm-5-3.md>), [google](<https://devfeed.tech/tags/google.md>), [groq](<https://devfeed.tech/tags/groq.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [ice-cubes](<https://devfeed.tech/tags/ice-cubes.md>), [inference-providers](<https://devfeed.tech/tags/inference-providers.md>), [inkling](<https://devfeed.tech/tags/inkling.md>), [jobbench](<https://devfeed.tech/tags/jobbench.md>), [kimi-k3](<https://devfeed.tech/tags/kimi-k3.md>), [long-running](<https://devfeed.tech/tags/long-running.md>), [m5-ultra](<https://devfeed.tech/tags/m5-ultra.md>), [mac-studio](<https://devfeed.tech/tags/mac-studio.md>), [mastodon](<https://devfeed.tech/tags/mastodon.md>), [mcp-atlas](<https://devfeed.tech/tags/mcp-atlas.md>), [meta](<https://devfeed.tech/tags/meta.md>), [meta-model-api](<https://devfeed.tech/tags/meta-model-api.md>), [meta-platforms](<https://devfeed.tech/tags/meta-platforms.md>), [meta-superintelligence-scaling-framework](<https://devfeed.tech/tags/meta-superintelligence-scaling-framework.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [mlx](<https://devfeed.tech/tags/mlx.md>), [model](<https://devfeed.tech/tags/model.md>), [model-context-protocol](<https://devfeed.tech/tags/model-context-protocol.md>), [muse-code](<https://devfeed.tech/tags/muse-code.md>), [muse-glimmer](<https://devfeed.tech/tags/muse-glimmer.md>), [muse-model-license](<https://devfeed.tech/tags/muse-model-license.md>), [muse-spark](<https://devfeed.tech/tags/muse-spark.md>), [muse-spark-1-2](<https://devfeed.tech/tags/muse-spark-1-2.md>), [muse-spark-open](<https://devfeed.tech/tags/muse-spark-open.md>), [news](<https://devfeed.tech/tags/news.md>), [open-weights](<https://devfeed.tech/tags/open-weights.md>), [platforms](<https://devfeed.tech/tags/platforms.md>), [release](<https://devfeed.tech/tags/release.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-3](<https://devfeed.tech/tags/top-story-3.md>)

### AI overview

Meta released open weights for Muse Spark Open, an 816-billion-parameter model designed for coding and long-running agent work. The report describes its multimodal inputs and million-token context window, tests involving coding tasks and tool use, and license terms for developers and hosted providers. Meta says it scored strongly on an agent tool-use benchmark, while Chinese open models scored higher on the coding benchmark cited.

### Source excerpt

Meta Platforms Inc. today released open weights for its Muse Spark artificial intelligence model, which had been proprietary since its April debut. The release, called Muse Spark Open, is built for coding and long-running agent work. Meta describes it as the most capable open-weight model to come out of a U.S. lab so far. Only [...] The post Meta releases 816B-parameter Muse Spark Open for coding and agent work appeared first on SiliconANGLE.

## Meta launches an open-weight AI model that runs on Macs

DevFeed: [Meta launches an open-weight AI model that runs on Macs](<https://devfeed.tech/articles/meta-launches-an-open-weight-ai-model-that-runs-on-macs-88311.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/10/15/meta-launches-an-open-weight-model-that-runs-on-macs/>)

Author: Lucas Ropek

Published: 2026-10-15T20:04:53Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [lunar foundation model](<https://devfeed.tech/topics/lunar-foundation-model.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [apple](<https://devfeed.tech/tags/apple.md>), [cost](<https://devfeed.tech/tags/cost.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [m5-ultra](<https://devfeed.tech/tags/m5-ultra.md>), [mac](<https://devfeed.tech/tags/mac.md>), [mac-studio](<https://devfeed.tech/tags/mac-studio.md>), [memory](<https://devfeed.tech/tags/memory.md>), [meta](<https://devfeed.tech/tags/meta.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [muse](<https://devfeed.tech/tags/muse.md>), [muse-code](<https://devfeed.tech/tags/muse-code.md>), [open-weight-ai](<https://devfeed.tech/tags/open-weight-ai.md>), [performance](<https://devfeed.tech/tags/performance.md>), [ram](<https://devfeed.tech/tags/ram.md>)

### AI overview

Meta launched Muse Spark Open, an open-weight multimodal model designed to run locally on Macs. It has 816 billion parameters and a one-million-token context window, and Meta says it is built for complex coding tasks and agentic workflows. Running it locally requires at least 1 terabyte of RAM; Meta also says it is available through several API platforms.

### Source excerpt

Meta and Apple have teamed up for the launch of the social media giant's newest model, Muse Spark Open, which will run locally on Apple silicon.

## Meta ships Muse Spark Open, its answer to China's best open-weight models

DevFeed: [Meta ships Muse Spark Open, its answer to China's best open-weight models](<https://devfeed.tech/articles/meta-ships-muse-spark-open-its-answer-to-china-s-best-open-weight-models-88244.md>)

Original publisher: [Read original article](<https://thenewstack.io/meta-muse-spark-open/>)

Author: Frederic Lardinois

Published: 2026-10-15T17:00:00Z

Content type: news

Language: en

Sources: [The New Stack](<https://devfeed.tech/sources/the-new-stack.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [code analysis](<https://devfeed.tech/topics/code-analysis.md>), [API Business Models](<https://devfeed.tech/topics/api-business-models.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [amazon-sagemaker](<https://devfeed.tech/tags/amazon-sagemaker.md>), [baseten](<https://devfeed.tech/tags/baseten.md>), [complex](<https://devfeed.tech/tags/complex.md>), [critical-infrastructure](<https://devfeed.tech/tags/critical-infrastructure.md>), [deepswe](<https://devfeed.tech/tags/deepswe.md>), [deepswe-1-1](<https://devfeed.tech/tags/deepswe-1-1.md>), [glm](<https://devfeed.tech/tags/glm.md>), [glm-5-3](<https://devfeed.tech/tags/glm-5-3.md>), [google-cloud](<https://devfeed.tech/tags/google-cloud.md>), [groq](<https://devfeed.tech/tags/groq.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [meta](<https://devfeed.tech/tags/meta.md>), [muse-spark](<https://devfeed.tech/tags/muse-spark.md>), [muse-spark-open](<https://devfeed.tech/tags/muse-spark-open.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>)

### AI overview

Meta launched Muse Spark Open, an 816B open-weight model with 42B active parameters, for API use and download under a license that includes commercial terms and user-interface attribution requirements. Meta reports strong agentic and tool-use benchmark results, while the article's cited coding benchmarks show it trailing Kimi K3 and GLM 5.3. Running the model locally would require two high-memory Mac Studios, according to Meta.

### Source excerpt

Meta on Thursday launched Muse Spark Open, a new 816B open model (with 42B active parameters) that the company argues The post Meta ships Muse Spark Open, its answer to China's best open-weight models appeared first on The New Stack.

## FLock.io Launches Qwen-powered, Open-Source Decision Model for Local Edge Deployment

DevFeed: [FLock.io Launches Qwen-powered, Open-Source Decision Model for Local Edge Deployment](<https://devfeed.tech/articles/flock-io-launches-qwen-powered-open-source-decision-model-for-local-edge-deployment-88337.md>)

Original publisher: [Read original article](<https://hackernoon.com/flockio-launches-qwen-powered-open-source-decision-model-for-local-edge-deployment?source=rss>)

Author: HackerNoon Press Releases

Published: 2026-10-15T16:09:04Z

Content type: news

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [Local AI](<https://devfeed.tech/topics/local-ai.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [qwen](<https://devfeed.tech/topics/qwen.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>)

Tags: [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [core](<https://devfeed.tech/tags/core.md>), [data-privacy](<https://devfeed.tech/tags/data-privacy.md>), [decentralize-ai](<https://devfeed.tech/tags/decentralize-ai.md>), [decision-model](<https://devfeed.tech/tags/decision-model.md>), [edge](<https://devfeed.tech/tags/edge.md>), [flock-io](<https://devfeed.tech/tags/flock-io.md>), [good-company](<https://devfeed.tech/tags/good-company.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [open-source-decision-model](<https://devfeed.tech/tags/open-source-decision-model.md>), [press-release](<https://devfeed.tech/tags/press-release.md>), [qwen](<https://devfeed.tech/tags/qwen.md>)

### AI overview

FLock.io and Oxford University researchers launched THIS/THAT, an open-source decision model built on Qwen3.5-2B for structured choices. The model is designed for local deployment on devices such as laptops and smartphones, with the stated aim of reducing latency and keeping sensitive data on-device. The article reports benchmark results, including 94.1% accuracy on 68 local-geometry questions and 87.8% success on laptop-based benchmarks.

### Source excerpt

FLock.io launches Qwen-powered, open-source decision model for local edge deployment.

## Open-weight models handle over 70% of gateway tokens and nearly one-quarter of spend

DevFeed: [Open-weight models handle over 70% of gateway tokens and nearly one-quarter of spend](<https://devfeed.tech/articles/open-weight-takes-71-5-of-tokens-and-23-of-spend-88217.md>)

Original publisher: [Read original article](<https://vercel.com/blog/ai-gateway-production-index-october-2026>)

Author: Harpreet Arora

Published: 2026-10-15T07:00:00Z

Content type: article

Language: en

Sources: [Vercel News](<https://devfeed.tech/sources/vercel-news.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [Claude](<https://devfeed.tech/topics/claude.md>)

Tags: [cost-per-token](<https://devfeed.tech/tags/cost-per-token.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [flash](<https://devfeed.tech/tags/flash.md>), [kimi-k3](<https://devfeed.tech/tags/kimi-k3.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [opus-5-5](<https://devfeed.tech/tags/opus-5-5.md>)

### AI overview

Vercel's October 2026 AI Gateway Production Index reports that open-weight models processed more than 70% of gateway tokens and nearly one-quarter of September spend, while overall gateway spend was about 13 times December's and average token prices were 47% lower. The report also compares model adoption, spending, provider locations, and launch performance using anonymized, aggregated gateway traffic.

### Source excerpt

AI Gateway Production Index -- October 2026 Every month, AI Gateway routes tens of trillions of tokens between production applications and AI labs. That traffic gives us a view of what AI usage actually looks like in today's enterprise, and we publish it here monthly. See the Production Index reports from July, August, and September. October 2026 summary Open-weight models now process more than seven in ten gateway tokens and account for nearly one in four dollars. Their share of spend has more than quadrupled since June. September's gateway spend was about 13 times December's, while the average price per token was 47% lower. Almost one in three teams changed their most-used lab in September, an index record. Nearly 70% of September's gateway spend on Chinese models went through US-based providers, up from less than 10% in December. Jev was the fastest-adopted model in AI Gateway history, reaching 35% of paid teams in its first two weeks, but it accounted for just 0.27% of gateway spend over that same period. Opus 5.5 took more than one in ten gateway dollars in its own first two weeks, the largest spend share of any September model launch. Frontier spend up 900%, spend share down 22% In August, open-weight models processed the majority of AI Gateway token volume for the first time. The money is now following. Open weight's share of spend has more than quadrupled over the last three months, from one in twenty gateway dollars in June to nearly one in four in September. Open weight's growth is coming from both ends of its price range, and it mirrors the wider gateway. The cheapest models drive token growth, while the expensive models drive spend. Led by Kimi K3, models priced at least four times the open-weight average cost per token accounted for about two-thirds of the increase in open-weight spend. In contrast, DeepSeek-v4.1-Flash alone processed nearly half of all September tokens but took just 2.6% of spend. Its price per token was about a fifth of the open-weight

## Meta Muse Spark Open is live on Together AI

DevFeed: [Meta Muse Spark Open is live on Together AI](<https://devfeed.tech/articles/meta-muse-spark-open-is-live-on-together-ai-88079.md>)

Original publisher: [Read original article](<https://www.together.ai/blog/meta-muse-spark-open>)

Published: 2026-10-15T00:00:00Z

Content type: release

Language: en

Sources: [Together.ai](<https://devfeed.tech/sources/together-ai.md>)

Topics: [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [agentic-coding](<https://devfeed.tech/tags/agentic-coding.md>), [agentic-engineering](<https://devfeed.tech/tags/agentic-engineering.md>), [agentic-software](<https://devfeed.tech/tags/agentic-software.md>), [ai](<https://devfeed.tech/tags/ai.md>), [context-window](<https://devfeed.tech/tags/context-window.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [long-context](<https://devfeed.tech/tags/long-context.md>), [meta](<https://devfeed.tech/tags/meta.md>), [ollama](<https://devfeed.tech/tags/ollama.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

Together AI announces that Meta Muse Spark Open, an open-weight model for agentic coding and other multimodal workflows, is available on its platform. The release highlights a 1M-token context window, long-running autonomous task support, computer use, browsing, and multimodal input. Early access customers Factory and Ollama describe evaluating it for tool calling and long-context work; Together AI lists serverless API pricing.

### Source excerpt

Meta Muse Spark Open, Meta's open-weight model for agentic coding, is live on Together AI with a 1M-token context window on serverless and dedicated infrastructure.

## AI tokenomics: Balancing open-weight and frontier models for optimal ROI

DevFeed: [AI tokenomics: Balancing open-weight and frontier models for optimal ROI](<https://devfeed.tech/articles/ai-tokenomics-balancing-open-weight-and-frontier-models-for-optimal-roi-86667.md>)

Original publisher: [Read original article](<https://www.redhat.com/en/blog/ai-tokenomics-balancing-open-weight-and-frontier-models-optimal-roi>)

Author: Ashish Nainwal

Published: 2026-10-14T00:00:00Z

Content type: article

Language: en

Sources: [Red Hat Blog](<https://devfeed.tech/sources/red-hat-blog.md>)

Topics: [LLM security](<https://devfeed.tech/topics/llm-security.md>), [shadow AI](<https://devfeed.tech/topics/shadow-ai.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Grafana](<https://devfeed.tech/topics/grafana.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Publish-subscribe pattern](<https://devfeed.tech/topics/pubsub.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-control-plane](<https://devfeed.tech/tags/ai-control-plane.md>), [ai-performance](<https://devfeed.tech/tags/ai-performance.md>), [ai-strategy](<https://devfeed.tech/tags/ai-strategy.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [api](<https://devfeed.tech/tags/api.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [balancing](<https://devfeed.tech/tags/balancing.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [better](<https://devfeed.tech/tags/better.md>), [board](<https://devfeed.tech/tags/board.md>), [frontier-models](<https://devfeed.tech/tags/frontier-models.md>), [models](<https://devfeed.tech/tags/models.md>), [open](<https://devfeed.tech/tags/open.md>), [roi](<https://devfeed.tech/tags/roi.md>)

### AI overview

The article argues that enterprises can manage rising AI costs by routing requests across open-weight and frontier models according to task complexity. It presents a directional cost model in which hybrid routing reduced five-year total cost of ownership by about 89% in the reference scenario, and says organizations should validate the estimates against their own usage and pricing.

### Source excerpt

Why airlines don't fly all-first-class planes and what it teaches us about AI strategyAirlines learned very early that they can't optimize a profitable route selling every seat as first class. Instead, the aircraft has multiple cabins, and the airline sizes each according to the route, the season, and the customers who use it. Operational flexibility is what makes the route economics work.Now picture the alternative: a plane with only first class seats available, which few of the passengers can afford. Either the planes fly with vacant seats, or the airline operates at a loss by offering a fir

## What is Decision 3.0? vLLM Semantic Router's New Open Decision Models

DevFeed: [What is Decision 3.0? vLLM Semantic Router's New Open Decision Models](<https://devfeed.tech/articles/what-is-decision-3-0-vllm-semantic-router-s-new-open-decision-models-85060.md>)

Original publisher: [Read original article](<https://www.marktechpost.com/2026/10/11/what-is-decision-3-0-vllm-semantic-routers-new-open-decision-models/>)

Author: Michal Sutter

Published: 2026-10-11T21:43:41Z

Content type: article

Language: en

Sources: [MarkTechPost](<https://devfeed.tech/sources/marktechpost.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-shorts](<https://devfeed.tech/tags/ai-shorts.md>), [apache-2-0](<https://devfeed.tech/tags/apache-2-0.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [decision-models](<https://devfeed.tech/tags/decision-models.md>), [editors-pick](<https://devfeed.tech/tags/editors-pick.md>), [evals](<https://devfeed.tech/tags/evals.md>), [for-devs](<https://devfeed.tech/tags/for-devs.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [new-releases](<https://devfeed.tech/tags/new-releases.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [semantic-router](<https://devfeed.tech/tags/semantic-router.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [staff](<https://devfeed.tech/tags/staff.md>), [tech-news](<https://devfeed.tech/tags/tech-news.md>), [technology](<https://devfeed.tech/tags/technology.md>), [uncategorized](<https://devfeed.tech/tags/uncategorized.md>), [vllm](<https://devfeed.tech/tags/vllm.md>)

### AI overview

vLLM Semantic Router released Decision 3.0, a family of five Apache-2.0 multimodal decision models ranging from 0.8B to 27B parameters. The models answer choice, yes/no, and score questions about text, JSON, and images by returning probabilities rather than generated text. The article reports internal benchmark results, latency measurements on one AMD Instinct MI325X GPU, and caveats about model rankings.

### Source excerpt

vLLM Semantic Router released Decision 3.0, 5 Apache-2.0 multimodal decision models from 0.8B to 27B. They answer choice, yes/no and score questions about text and images in one call, without generating text. The post What is Decision 3.0? vLLM Semantic Router's New Open Decision Models appeared first on MarkTechPost.

## OpenAI revenue falls short, models play hopscotch and Trump cracks down on tech green cards

DevFeed: [OpenAI revenue falls short, models play hopscotch and Trump cracks down on tech green cards](<https://devfeed.tech/articles/openai-revenue-falls-short-models-play-hopscotch-and-trump-cracks-down-on-tech-green-cards-84061.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/10/09/openai-revenue-falls-short-models-play-hopscotch-and-trump-cracks-down-on-tech-green-cards/>)

Author: Robert Hof

Published: 2026-10-09T14:00:29Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [OpenAI](<https://devfeed.tech/topics/openai.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [alphabet](<https://devfeed.tech/tags/alphabet.md>), [anaconda](<https://devfeed.tech/tags/anaconda.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [atlassian](<https://devfeed.tech/tags/atlassian.md>), [bret-taylor](<https://devfeed.tech/tags/bret-taylor.md>), [company](<https://devfeed.tech/tags/company.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cyber](<https://devfeed.tech/tags/cyber.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [google](<https://devfeed.tech/tags/google.md>), [google-cloud](<https://devfeed.tech/tags/google-cloud.md>), [ibm](<https://devfeed.tech/tags/ibm.md>), [infra](<https://devfeed.tech/tags/infra.md>), [investors](<https://devfeed.tech/tags/investors.md>), [isometric-labs](<https://devfeed.tech/tags/isometric-labs.md>), [meta](<https://devfeed.tech/tags/meta.md>), [models](<https://devfeed.tech/tags/models.md>), [news](<https://devfeed.tech/tags/news.md>), [openai](<https://devfeed.tech/tags/openai.md>), [ptc](<https://devfeed.tech/tags/ptc.md>), [red-hat](<https://devfeed.tech/tags/red-hat.md>), [sailpoint](<https://devfeed.tech/tags/sailpoint.md>), [sailpoint-navigate](<https://devfeed.tech/tags/sailpoint-navigate.md>), [schneider-electric](<https://devfeed.tech/tags/schneider-electric.md>), [security](<https://devfeed.tech/tags/security.md>), [sierra-technologies](<https://devfeed.tech/tags/sierra-technologies.md>), [tanium](<https://devfeed.tech/tags/tanium.md>), [tech](<https://devfeed.tech/tags/tech.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-2](<https://devfeed.tech/tags/top-story-2.md>)

### AI overview

This weekly enterprise and emerging technology roundup reports that OpenAI's revenue estimate was $18 billion below its earlier stated figure, alongside new AI models and products, investment activity, and a U.S. policy action affecting technology companies' employee green card applications. It also surveys developments in quantum computing, AI research, enterprise tools, and technology leadership.

### Source excerpt

I'm not going to call the bursting of artificial intelligence bubble yet, not with all that money still pouring into every hardware and software company with AI in their pitch decks. But we got a little taste this week. OpenAI told investors this week that it actually had $18 billion less revenue than the $68 [...] The post OpenAI revenue falls short, models play hopscotch and Trump cracks down on tech green cards appeared first on SiliconANGLE.

## Cloudflare Open Sources Decision Models for AI Agents

DevFeed: [Cloudflare Open Sources Decision Models for AI Agents](<https://devfeed.tech/articles/cloudflare-open-sources-decision-models-for-ai-agents-77059.md>)

Original publisher: [Read original article](<https://www.infoq.com/news/2026/10/clef-decision-models/>)

Author: Renato Losio

Published: 2026-10-08T06:54:00Z

Content type: news

Language: en

Sources: [InfoQ](<https://devfeed.tech/sources/infoq.md>)

Topics: [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [Cloudflare](<https://devfeed.tech/topics/cloudflare.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Cloudflare Workers](<https://devfeed.tech/topics/cloudflare-workers.md>)

Tags: [agentic-ai-architecture](<https://devfeed.tech/tags/agentic-ai-architecture.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-ml-data-engineering](<https://devfeed.tech/tags/ai-ml-data-engineering.md>), [architecture-design](<https://devfeed.tech/tags/architecture-design.md>), [clef-decision-models](<https://devfeed.tech/tags/clef-decision-models.md>), [cloudflare](<https://devfeed.tech/tags/cloudflare.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [ml-data-engineering](<https://devfeed.tech/tags/ml-data-engineering.md>), [model-inference](<https://devfeed.tech/tags/model-inference.md>), [models](<https://devfeed.tech/tags/models.md>), [news](<https://devfeed.tech/tags/news.md>), [open](<https://devfeed.tech/tags/open.md>), [open-weight-ai](<https://devfeed.tech/tags/open-weight-ai.md>)

### AI overview

Cloudflare announced Clef, a set of open-weight multimodal models that classify inputs into predefined options with probabilities, enabling AI agents to make decisions such as routing or escalating support requests. The 9B Clef-Flash is aimed at latency-sensitive tasks, and Cloudflare offers model access through Workers AI and downloadable weights, with fine-tuning options for specific workloads.

### Source excerpt

During its recent "Birthday Week", Cloudflare announced Clef, a set of open-weight AI models designed to choose between predefined options rather than generate text. Cloudflare released 9B- and 27B-parameter models, along with a platform for adapting them to specific decision-making tasks. By Renato Losio

## Automatic local AI model routing with NVIDIA Switchyard

DevFeed: [Automatic local AI model routing with NVIDIA Switchyard](<https://devfeed.tech/articles/automatic-local-ai-model-routing-with-nvidia-switchyard-65109.md>)

Original publisher: [Read original article](<https://technotim.com/posts/switchyard-local-ai-routing/>)

Author: Techno Tim

Published: 2026-10-05T13:00:00Z

Content type: article

Language: en

Sources: [Techno Tim](<https://devfeed.tech/sources/techno-tim.md>)

Topics: [Model Routing](<https://devfeed.tech/topics/model-routing.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [gpt-oss](<https://devfeed.tech/topics/gpt-oss.md>), [CUDA Graphs](<https://devfeed.tech/topics/cuda-graphs.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Scapy](<https://devfeed.tech/topics/scapy.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [automation](<https://devfeed.tech/tags/automation.md>), [code](<https://devfeed.tech/tags/code.md>), [homelab](<https://devfeed.tech/tags/homelab.md>), [local-ai](<https://devfeed.tech/tags/local-ai.md>), [local-models](<https://devfeed.tech/tags/local-models.md>), [model](<https://devfeed.tech/tags/model.md>), [model-routing](<https://devfeed.tech/tags/model-routing.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>)

### AI overview

The article evaluates NVIDIA Switchyard for routing coding-agent requests between a faster local Nemotron model and a more capable DeepSeek model. It describes how increasing Nemotron's context window reduced fallback traffic, and reports that in one automated test-suite run, 87.7% of model selections stayed on Nemotron. The author also covers configuration, monitoring, benchmark limitations, and safeguards for a local deployment.

### Source excerpt

I've been running a few local models, and I don't really want to choose one before every request. A faster model might handle most of a coding task just fine, but sometimes the agent could run into a failing test or start chasing an error that needs a little more reasoning. I could send everything to the most capable model, but I'd also be waiting on it for work the faster one could have handle...

[Next page](<https://devfeed.tech/topics/gpt-oss.md?cursor=WyIyMDI2LTEwLTA1VDEzOjAwOjAwKzAwOjAwIiwgImQzZmZhMTM1LWZhNGItNDc2NC1iOGQxLWJjMjAwYzQ0ZmUyMCJd>)