# deepseek Published articles for deepseek. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard DevFeed: [HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard]() Original publisher: [Read original article]() Author: Brian Beeler Published: 2026-09-15T23:15:57Z Content type: article Language: en Sources: [StorageReview.com]() Topics: [Local AI](), [Hardware](), [GPU](), [cpu](), [Large Language Model]() Tags: [consumer](), [cpu](), [deepseek](), [dgx-spark](), [gemma](), [gpu](), [hardware](), [local-ai](), [models](), [ollama](), [windows](), [workstation]() ### AI overview StorageReview previews HP's pre-production ZBook Ultra G3a 16, a local AI laptop with 192GB of unified memory and up to 160GB assignable to its integrated GPU. The article examines its hardware and planned testing while noting that shipping-hardware benchmarks are not yet available. ### Source excerpt HP's ZBook Ultra G1a 14 holds the Best for Large Models spot on our Best Laptops for Local AI leaderboard because its 128GB of unified memory, 96GB of it assignable to the GPU, loaded models no discrete-GPU laptop could touch. The new HP ZBook Ultra G3a 16 raises that pool to 192GB with up to The post HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard appeared first on StorageReview.com. ## Accelerating Dropless MoE Training in JAX with NVIDIA Transformer Engine DevFeed: [Accelerating Dropless MoE Training in JAX with NVIDIA Transformer Engine]() Original publisher: [Read original article]() Author: Tanya Lenz Published: 2026-09-14T16:39:15Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Mixture of Experts (MoE)](), [Artificial Intelligence](), [GPU](), [Large Language Model](), [Nvidia](), [Routing (disambiguation)](), [GB200](), [Network](), [Python](), [deepseek](), [mixtral](), [qwen]() Tags: [agentic-ai-generative-ai](), [ai](), [communication](), [deepseek](), [developer-tools-techniques](), [gb200](), [gpu](), [mixtral](), [mixture-of-experts-moe](), [mlops](), [moe](), [networks](), [nvfp4](), [nvidia](), [performance](), [python](), [qwen](), [router](), [routing](), [tensors](), [token](), [tokens](), [training-ai-models]() ### AI overview This article explains how NVIDIA Transformer Engine and JAX kernel optimizations accelerate dropless Mixture of Experts (MoE) training. It describes bottlenecks from token routing, expert dispatch and gathering, all-to-all communication, and ragged expert matrix operations. In DeepSeek-V3 training on NVIDIA GB200, the optimized approach increased performance from 103 to 1,068 TFLOPS per GPU, a 10.4x improvement. ### Source excerpt Mixture of experts (MoE) has become one of the defining architectural trends in large-scale AI model training. DeepSeek, Qwen, and Mixtral are examples of MoE... ## What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent DevFeed: [What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent]() Original publisher: [Read original article]() Author: Rudrendu Paul Published: 2026-09-11T15:07:18Z Content type: tutorial Language: en Sources: [freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More]() Topics: [AI Bots](), [AI-assisted coding]() Tags: [agent](), [ai-agents](), [architecture](), [artificial-intelligence](), [claude-code](), [deepseek](), [developer-tools](), [llm](), [machine-learning](), [mcp](), [memory](), [python](), [sandbox](), [software-engineering](), [tools]() ### AI overview An explainer and hands-on guide to agent harnesses: the runtime infrastructure around an LLM that manages model calls, tool execution, memory, and filesystem sandboxing. It compares popular harnesses and introduces a small Python implementation. ### Source excerpt On August 13, 2026, DeepSeek published a GitHub repository called deepseek-harness. Within two days, it had passed 95,386 stars and 8,826 forks (a vanity metric on its own, but a spike this fast signa ## DeepSeek's new model sets a template for powerful LLMs that run lean DevFeed: [DeepSeek's new model sets a template for powerful LLMs that run lean]() Original publisher: [Read original article]() Author: Tobias Mann Published: 2026-09-11T07:15:00Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [deepseek](), [Large Language Model](), [Caching](), [Latency](), [Google]() Tags: [ai-and-ml](), [cache](), [compute](), [cost](), [datacenter](), [deepseek](), [embedding](), [flash](), [google](), [gpu](), [latency](), [llm](), [llms](), [model](), [release]() ### AI overview DeepSeek V4.1 Flash is a larger LLM whose architectural changes aim to reduce serving memory and compute needs. The article highlights lower KV-cache consumption, improved prompt processing, and an N-gram-based conditional memory module. ### Source excerpt DeepSeek V4.1 Flash proves that just because you build a bigger model doesn't mean you need more GPUs to serve it ## Hundreds of AI agents helped PaperCut attacker hit 395+ orgs, and some went off script DevFeed: [Hundreds of AI agents helped PaperCut attacker hit 395+ orgs, and some went off script]() Original publisher: [Read original article]() Author: Jessica Lyons Published: 2026-09-10T18:49:43Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [AI Bots]() Tags: [agents](), [ai](), [ai-agents](), [ai-and-ml](), [cyber-crime](), [deepseek](), [openai](), [security]() ### AI overview Hundreds of AI agents reportedly helped a PaperCut attacker target more than 395 organizations. The agents reportedly disregarded an instruction not to target CIS organizations. ### Source excerpt Human operator: don't touch CIS orgs. AI agents: look a squirrel! ## Detect and disrupt AI-themed attacks with Microsoft Defender DevFeed: [Detect and disrupt AI-themed attacks with Microsoft Defender]() Original publisher: [Read original article]() Author: Rob Lefferts Published: 2026-09-10T16:00:00Z Content type: article Language: en Sources: [Microsoft Security Blog]() Topics: [Malware](), [Threat Research](), [spoofing](), [Microsoft]() Tags: [adversary-in-the-middle-aitm](), [ai](), [attacks](), [chatgpt](), [claude](), [copilot](), [credential-theft](), [deepseek](), [defender](), [github](), [malware](), [microsoft](), [phishing](), [security](), [social-engineering]() ### AI overview Microsoft describes AI-themed phishing, malvertising, credential theft, and malware campaigns that impersonate popular AI services and tools. It argues that attackers are exploiting trust and urgency around AI brands rather than compromising the referenced services. ### Source excerpt See how Microsoft Defender detects and disrupts AI-themed phishing, malware, and multi-stage attacks across the attack chain. The post Detect and disrupt AI-themed attacks with Microsoft Defender appeared first on Microsoft Security Blog. ## DeepSeek V4.1 Flash now available on AI Gateway DevFeed: [DeepSeek V4.1 Flash now available on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-09-09T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [deepseek](), [Artificial Intelligence](), [API](), [Language models](), [AI-assisted coding](), [Claude Code](), [codex](), [cursor]() Tags: [ai](), [ai-gateway](), [api](), [caching](), [codex](), [context-window](), [cursor](), [deepseek](), [responses](), [tool](), [vercel](), [vision]() ### AI overview DeepSeek V4.1 Flash is now available through Vercel's AI Gateway, offering native image understanding, a 1 million token context window, responses up to 384,000 tokens, reasoning, tool use, and prompt caching. Developers can use it in Claude Code, Codex, Cursor, and other coding agents with the model name deepseek/deepseek-v4.1-flash. ### Source excerpt DeepSeek V4.1 Flash is now available on AI Gateway with native image understanding. V4.1 Flash has vision support and accepts text and images in the same request, so you can ask questions about screenshots, read charts, and extract information from visual content. The model has a 1 million token context window and supports responses up to 384,000 tokens, along with reasoning, tool use, and prompt caching. Its new architecture processes input and generates output with separate components, reducing the active computation needed for each stage. Use deepseek/deepseek-v4.1-flash as the model name: To use it in Claude Code, Codex, Cursor, and more, install the latest Vercel CLI and run setup: Then select deepseek/deepseek-v4.1-flash in the agent. See the coding agents guide for details. Try DeepSeek V4.1 Flash in the model playground. AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. It includes built-in custom reporting, Zero Data Retention support, budgets for API keys, routing rules, and more. AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests. You can view all language models available on AI Gateway. Read more ## Optimize vLLM speculative decoding with FastMTP heads DevFeed: [Optimize vLLM speculative decoding with FastMTP heads]() Original publisher: [Read original article]() Author: Rahul Tuli Published: 2026-09-08T14:20:16Z Content type: article Language: en Sources: [Red Hat](), [Red Hat Developer]() Topics: [vllm](), [Inference](), [Large Language Model](), [Fine-tuning](), [model-deployment](), [deepseek](), [qwen]() Tags: [ai-inference](), [artificial-intelligence](), [data](), [deepseek](), [developer-tools](), [fine-tuning](), [hardware](), [inference](), [large-language-model](), [production](), [qwen](), [tokens](), [vllm]() ### AI overview This article explains how FastMTP-style fine-tuning improves vLLM speculative decoding. It describes using native multi-token prediction heads as speculators, adapting a single head for recursive multi-step drafting, extracting weights from verifier checkpoints, and producing vLLM-ready checkpoints without training from scratch. ### Source excerpt Autoregressive decoding makes large language model (LLM) inference memory-bandwidth bound: every token needs 1 full forward pass over billions of parameters, so the hardware spends most of its time moving weights rather than computing. MTP is a training objective: models like the DeepSeek and Qwen families learn to predict several future tokens at each position, which improves their data efficiency and quality. The post Optimize vLLM speculative decoding with FastMTP heads appeared first on Red Hat Developer. ## The Agentic Analytics Benchmark: Measuring model accuracy and efficiency in analytical agents DevFeed: [The Agentic Analytics Benchmark: Measuring model accuracy and efficiency in analytical agents]() Original publisher: [Read original article]() Author: Eduardo Vellasques; Al Brown Published: 2026-09-01T19:15:00Z Content type: article Language: en Sources: [ClickHouse Blog]() Topics: [LLM evaluation / benchmarking](), [benchmarking](), [text2sql](), [Databases](), [Claude](), [deepseek]() Tags: [agentic](), [agents](), [benchmark](), [claude](), [cost](), [deepseek](), [model](), [speed](), [sql]() ### AI overview The article introduces data-agent-mnist, an open benchmark harness for evaluating analytics agents against a data warehouse. It compares 29 models on correctness, cost, speed, and number of turns, and contrasts agentic analytics with conventional text-to-SQL evaluation. ### Source excerpt We took 201 real analytics questions from our production data warehouse, benchmarked 29 models on correctness, cost, and speed, and released an open harness so you can run the same test on your own. ## Hot Chips 2026: Intel's Crescent Island DevFeed: [Hot Chips 2026: Intel's Crescent Island]() Original publisher: [Read original article]() Author: George Cozma Published: 2026-08-27T05:12:31Z Content type: article Language: en Sources: [Chips and Cheese]() Topics: [intel](), [GPU](), [Artificial Intelligence](), [datacenter](), [Blackwell](), [deepseek]() Tags: [2026](), [ai](), [amd](), [blackwell](), [datacenter](), [deepseek](), [gpu](), [intel]() ### AI overview The article discusses Intel's upcoming Crescent Island datacenter GPU, focusing on its LPDDR5X-based memory subsystem, reported capacity of up to 480 GB, undisclosed memory bandwidth, Xe3P architecture, and full-rate FP64 support. It also compares aspects of the design with competing GPUs. ### Source excerpt Hello you fine Internet folks, ## Jalapeño's first results show industry-leading speed and efficiency in AI inference DevFeed: [Jalapeño's first results show industry-leading speed and efficiency in AI inference]() Original publisher: [Read original article]() Published: 2026-08-25T07:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Artificial Intelligence](), [Inference](), [Inference Performance](), [OpenAI](), [Hardware](), [Benchmark](), [gpt-oss](), [deepseek]() Tags: [ai](), [benchmark](), [deepseek](), [efficiency](), [engineering](), [gpt-oss](), [hardware](), [inference](), [latency](), [models](), [openai](), [performance](), [software](), [speed](), [systems]() ### AI overview OpenAI reports initial results for Jalapeño, its custom inference chip. The article says the chip delivers higher throughput, lower end-to-end latency, and greater AI work per watt across GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, based on tests using the InferenceX benchmark. ### Source excerpt Jalapeño is a custom inference chip from OpenAI that delivers faster, more power-efficient AI inference, with higher throughput and lower latency for modern models. ## DeepSeek V4 Flash Vision Experimental now available on AI Gateway DevFeed: [DeepSeek V4 Flash Vision Experimental now available on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-08-21T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [deepseek](), [Claude](), [Artificial Intelligence](), [Language models](), [Inference](), [Caching](), [Vercel](), [Claude Code](), [codex](), [cursor]() Tags: [ai](), [ai-gateway](), [caching](), [claude](), [claude-code](), [codex](), [coding-agents](), [cursor](), [deepseek](), [experimental](), [inference](), [language-models](), [playground](), [pricing](), [reasoning](), [tool](), [vercel](), [vision]() ### AI overview DeepSeek V4 Flash Vision Experimental is now available through Vercel AI Gateway. The model accepts images with text, supports tasks such as image description, screenshot text extraction, and chart analysis, and retains tool use, reasoning, and caching. The article also documents supported image formats, experimental-release caveats, coding-agent setup, playground access, and AI Gateway pricing. ### Source excerpt DeepSeek V4 Flash with vision is now available on AI Gateway. This model is an experimental version that accepts images alongside text. You can ask it to describe a picture, read text out of a screenshot, or work through a chart in the same request as your prompt. DeepSeek V4 Flash Vision Experimental now available on AI Gateway. Tool use, reasoning, and caching all work the same as before. Use deepseek/deepseek-v4-flash-vision-exp to get started: Images can be JPEG, PNG, GIF, or WebP. The format is read from the file's own bytes rather than its name or the mediaType you declare, so a mislabeled file still goes through. The -exp in the model ID marks this as an experimental release. Expect behavior to change, and keep a fallback model configured if it's on a production path. To use it in a coding agent, run vercel ai-gateway coding-agents setup to connect agents like Claude Code, Codex, OpenCode, Cursor, and Pi, then select deepseek/deepseek-v4-flash-vision-exp inside the agent. Try Deepseek V4 Flash Vision Experimental in the model playground. AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including BYOK requests. View all language models on AI Gateway to see more. Read more ## How Much Memory Does Your Agent Actually Need? DevFeed: [How Much Memory Does Your Agent Actually Need?]() Original publisher: [Read original article]() Author: Vatche Isahagian; Gaodan Fang; Jayaram Radhakrishnan; Punleuk Oum; Ashwath Vaithinathan Aravindan; Evelyn Duesterwald; G Thomas; Vinod Muthusamy; Merve Unuvar; Ayhan Sebin Published: 2026-08-18T18:09:38Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [Benchmark](), [Inference](), [long-context](), [gpt-oss](), [deepseek](), [Mixture of Experts (MoE)]() Tags: [agent](), [agentic](), [caching](), [deepseek](), [evaluation](), [gpt-oss](), [inference](), [memory](), [model](), [models](), [moe](), [performance](), [production](), [tokens]() ### AI overview ALTK-Evolve equips agents with agentic memory by distilling reusable guidelines from past trajectories and injecting them at inference time without weight updates or human annotation. An evaluation across eight models found that the effective amount of memory varies by model: stronger models benefited from full guideline sets, weaker models from compact cores plus task-specific retrieval, and saturated models showed no measurable improvement. For gpt-oss-120b, selective retrieval improved task completion by 16.1 percentage points with only a 5% token increase. ### Source excerpt Equipping an agent with agentic memory sounds simple: distill lessons from its past work, put them back in context, and more experience should mean better performance. It doesn't always work that way. When we scaled the evaluation to eight models -- from a 30B dense model to frontier proprietary systems -- one finding stood out: Agentic memory is not a feature you switch on. It's a dose you calibrate to the model. ## Agent Night panel recap: AI spending, open-source models, and the changing agent economy DevFeed: [Agent Night panel recap: AI spending, open-source models, and the changing agent economy]() Original publisher: [Read original article]() Author: WorkOS Published: 2026-08-17T18:34:23Z Content type: article Language: en Sources: [WorkOS Blog]() Topics: [Artificial Intelligence](), [Open Source](), [deepseek](), [OpenAI]() Tags: [ai](), [anthropic](), [deepseek](), [engineering](), [harness](), [open-source](), [openai](), [recap]() ### AI overview A recap of WorkOS's Agent Night panel, where Jaya Gupta, Flo Crivello, and swyx discussed increased spending on AI tooling, progress in open-source models, and changing economics for AI agents. The article presents panelists' claims and observations, including Flo Crivello's view that AGI has arrived and reports of companies moving from OpenAI and Anthropic to open-source models. ### Source excerpt Jaya Gupta, Flo Crivello, and swyx on stage at Agent Night: 5x AI spend, open source getting real, harness engineering, skills, and December predictions. ## How to switch models in Claude Code with one command DevFeed: [How to switch models in Claude Code with one command]() Original publisher: [Read original article]() Author: WorkOS Published: 2026-08-14T14:04:26Z Content type: tutorial Language: en Sources: [WorkOS Blog]() Topics: [Claude Code](), [Protocol (disambiguation)](), [Bash](), [Command-line interface](), [API](), [SDKs](), [Vercel](), [deepseek]() Tags: [api](), [claude-code](), [command-line](), [commands](), [deepseek](), [how-to](), [vercel]() ### AI overview A tutorial explains how to switch models in Claude Code by using a shell wrapper and environment variables with Anthropic-compatible gateways such as Vercel AI Gateway and OpenRouter. It highlights configuration details, including clearing conflicting credentials and correctly setting empty variables, and notes that the Claude Agent SDK inherits the same setup. ### Source excerpt Claude Code speaks Anthropic's Messages protocol and so do the major gateways. A small shell wrapper points it at any model, with live rate cards to prove it. ## Route AI Agent Workloads Across Models with NVIDIA NeMo Switchyard DevFeed: [Route AI Agent Workloads Across Models with NVIDIA NeMo Switchyard]() Original publisher: [Read original article]() Author: Michelle Horton Published: 2026-08-11T13:00:00Z Content type: tutorial Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [AI Bots]() Tags: [agent](), [agentic](), [agentic-ai-generative-ai](), [agents](), [ai](), [ai-agent](), [applications](), [benchmark](), [classification](), [computer-use](), [cost](), [deepseek](), [developer-tools-techniques](), [developers](), [engineering](), [featured](), [latency](), [llms](), [math](), [model](), [models](), [nemo](), [nvidia](), [reasoning](), [top-stories]() ### AI overview The article explains how NVIDIA NeMo Switchyard routes AI-agent tasks to different models according to task requirements, capabilities, cost, and latency. ### Source excerpt Building an AI agent does not end with choosing a single model. Each model has its own strengths, weaknesses, and cost profile, which can shift from one... ## DeepSeek overtakes Google on volume, cost per token falls 13.6% DevFeed: [DeepSeek overtakes Google on volume, cost per token falls 13.6%]() Original publisher: [Read original article]() Author: Harpreet Arora Published: 2026-08-11T04:00:00Z Content type: news Language: en Sources: [Vercel News]() Topics: [AI Chat]() Tags: [agent](), [ai](), [ai-gateway](), [anthropic](), [applications](), [claude](), [cost](), [data](), [deepseek](), [enterprise](), [google](), [gpt](), [launch](), [leaderboards](), [media](), [models](), [openai](), [production](), [video](), [work]() ### AI overview The August AI Gateway Production Index reports that token prices fell 13.6% in July while token use and spending grew. DeepSeek surpassed Google for second place by token volume, while Kimi K3 and other open-weight models gained usage and spending. ### Source excerpt AI Gateway Production Index -- August 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 May, June, and July. August 2026 Summary The August index reports on AI Gateway data collected through July 2026. The average price paid per token fell 13.6% in July, after rising almost 20% in May and holding steady in June. Token consumption grew so quickly that even with 37% growth in spend, cost per token saw a double-digit drop. DeepSeek became the second-largest lab by token volume, now running more than twice Google's volume. Anthropic collected 65% of gateway spending on 30% of token volume, at 4.4 times the average price of every other lab's tokens. Both media leaderboards changed hands. Google's Nano Banana took the lead in image volume from OpenAI's GPT Image, and ByteDance's Seedance led video in both volume and dollars. Kimi K3 launched into agent work Moonshot released Kimi K3 on July 16. Like Z.ai's GLM 5.2 released in June, it is built for long-horizon agent work, so its usage was heavy from the start at about twelve times the tokens per request of its predecessor, K2.5. It scaled quickly. K3's daily volume tripled between launch week and the final week of July, and by month end its requests were as heavy as Claude Opus 4.8's. On the last full day of July it ranked eighth on the gateway by token volume. The demand was new rather than diverted. K3 processed nearly two-thirds of all Kimi tokens within two weeks and 82% by the final week, while the rest of the family's volume fell only slightly. Open weight's share of gateway spend more than doubled in July to 8.6%. More than 90% of that growth is Moonshot and Z.ai. Moonshot's share of total gateway spend quadrupled, to 2.3%. Cheap open-weight models have been taking volume for months without taking reve ## 🗓 This Week In AI Research (25-31 July 26) DevFeed: [🗓 This Week In AI Research (25-31 July 26)]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-08-07T01:00:43Z Content type: article Language: en Sources: [Into AI]() Topics: [releases](), [Machine learning](), [deepseek](), [foundation-models](), [Mixture of Experts (MoE)](), [lora](), [codex]() Tags: [ai](), [deep-learning](), [deepseek](), [lora](), [ml](), [moe](), [performance](), [releases](), [research]() ### AI overview A weekly roundup of AI research and releases covering DeepSeek-V4-Flash-0731, the Pangram 4 AI-text classification model, the OpenMLE system and its Frontis-MA1-35B agent, and the Metis memory foundation model. ### Source excerpt The top 10 AI research papers and releases this week. ## Baseten on Hugging Face Inference Providers 🔥 DevFeed: [Baseten on Hugging Face Inference Providers 🔥]() Original publisher: [Read original article]() Author: Alex Ker; Roland Crosby; Sid Shanker; Johan; Célina Hanouti; Simon Brandeis; Lucain Pouget; merve Published: 2026-08-06T00:00:00Z Content type: news Language: en Sources: [Hugging Face - Blog]() Topics: [AI Chat](), [SDKs]() Tags: [baseten](), [deepseek](), [enterprise](), [hub](), [hugging-face](), [inference](), [inference-providers](), [llms](), [partnerships](), [sdks](), [text-generation]() ### AI overview Baseten is now an Inference Provider on the Hugging Face Hub, offering conversational and text-generation access to supported open-weight LLMs. The integration is available through Hugging Face Python and JavaScript SDKs, with custom-key and Hugging Face-routed request modes. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## DeepSeek V4 Flash is 90% off through Novita on AI Gateway DevFeed: [DeepSeek V4 Flash is 90% off through Novita on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-08-04T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [deepseek](), [API](), [API keys](), [Vercel](), [dashboards]() Tags: [ai](), [api](), [customers](), [deepseek](), [flash](), [model](), [playground](), [vercel]() ### AI overview DeepSeek v4 Flash is available at a 90% discount through Novita on Vercel's AI Gateway for Vercel Pro customers through August 11. Users can configure the specified DeepSeek model, prioritize Novita with fallback providers, create an API key, or test the model in the playground. ### Source excerpt DeepSeek v4 Flash is 90% off on AI Gateway when you route to Novita. The discount is available to Vercel Pro customers through August 11. To get the discounted rate, set the model to deepseek/deepseek-v4-flash or deepseek/deepseek-v4-flash-0731 and put Novita first with the order option: If Novita can't serve the request, it falls back to other providers at the standard rate. After August 11, the model stays available at standard rates with no markup. Create an API key in the AI Gateway section of your dashboard, or try the model first in its playground. Read more ## DeepSeek V4 Flash now runs updated weights on AI Gateway DevFeed: [DeepSeek V4 Flash now runs updated weights on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-07-31T07:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [AI, ML & Data Engineering](), [vercel ai sdk](), [AI-assisted coding]() Tags: [agent](), [agentic](), [ai-gateway](), [coding](), [deepseek](), [flash](), [inference](), [model](), [vercel]() ### AI overview DeepSeek V4 Flash now uses updated weights by default through AI Gateway, improving its Terminal-Bench score without requiring a model ID or code change. ### Source excerpt DeepSeek V4 Flash now runs on updated weights by default on AI Gateway, with notably stronger agentic capabilities. On Terminal-Bench, it scores 82.7, up 25.8 points from 56.9 in the April preview. Requests to deepseek/deepseek-v4-flash pick up the new weights automatically, with no change to the model ID or your code. For now, DeepSeek is the only provider serving the updated weights. Other providers, including ones with Zero Data Retention, are coming next week. To use the updated DeepSeek V4 Flash, set model to deepseek/deepseek-v4-flash in the AI SDK. AI Gateway will route to providers with the new weights by default: To run V4 Flash in a coding agent, use vercel ai-gateway coding-agents setup to connect your agents to AI Gateway, then select deepseek/deepseek-v4-flash in the agent's model configuration. See the coding agents guide. AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests. Learn more about AI Gateway, view the AI Gateway model leaderboard or try it in our model playground. Read more ## The Rise of Open-Weight AI Models and the Challenge to Commercial Frontier Labs DevFeed: [The Rise of Open-Weight AI Models and the Challenge to Commercial Frontier Labs]() Original publisher: [Read original article]() Author: Colin Eberhardt Published: 2026-07-27T00:00:00Z Content type: opinion Language: en Sources: [Scott Logic]() Topics: [Artificial Intelligence](), [AI Models](), [Open Source](), [OpenAI](), [anthropic](), [deepseek](), [ChatGPT](), [llama](), [Reinforcement learning](), [Transformer]() Tags: [2026](), [ai](), [ai-models](), [anthropic](), [artificial-intelligence](), [chatgpt](), [china](), [cost](), [deepseek](), [llama](), [open-source](), [openai](), [performance](), [privacy]() ### AI overview This article traces the development of commercial and open-weight AI models from OpenAI's founding through the rise of models predominantly emanating from China. It argues that open-weight models have narrowed or closed the capability gap with frontier labs and may offer comparable or better performance at lower cost, with greater ownership and data privacy. ### Source excerpt Chinese-led open-weight AI models have closed the capability gap with frontier labs like OpenAI and Anthropic, challenging the assumption that massive investments guarantee lasting competitive advantage. By 2026, open-weight alternatives offer equivalent or superior performance at a fraction of the cost, along with greater ownership and data privacy, fundamentally disrupting the commercial AI landscape. ## Setting a World Record for MoE Pre-Training on NVIDIA GB300 NVL72 DevFeed: [Setting a World Record for MoE Pre-Training on NVIDIA GB300 NVL72]() Original publisher: [Read original article]() Author: Kirthi Devleker Published: 2026-07-21T18:30:00Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Mixture of Experts (MoE)](), [GPU](), [Nvidia](), [AI Platform](), [NCCL](), [networking](), [deepseek]() Tags: [agentic-ai-generative-ai](), [ai](), [ai-platform](), [ai-training](), [collective](), [communication](), [compute](), [data-center-cloud](), [deepseek](), [developer-tools-techniques](), [efficiency](), [featured](), [frontier-model](), [gb300-nvl72](), [gpu](), [infrastructure](), [latency](), [llm-techniques](), [low-latency](), [megatron](), [mixture-of-experts-moe](), [moe](), [nvidia](), [nvl72](), [performance](), [top-stories](), [train](), [training-ai-models]() ### AI overview The article explains how NVIDIA GB300 NVL72 achieved a world record for DeepSeek-V3 671B mixture-of-experts pre-training. It focuses on the communication demands of MoE models, including all-to-all traffic between GPUs, and the need for tightly coupled scale-up and predictable scale-out networking to sustain delivered training performance. ### Source excerpt Frontier model pre-training has converged on mixture of experts (MoE), which is fundamentally changing what limits large-scale AI training. As compute per token... ## Open-weight models surge to 29% of volume, price per token flattens DevFeed: [Open-weight models surge to 29% of volume, price per token flattens]() Original publisher: [Read original article]() Author: Harpreet Arora Published: 2026-07-13T07:00:00Z Content type: article Language: en Sources: [Vercel News]() Topics: [Artificial Intelligence](), [data](), [anthropic](), [deepseek](), [OpenAI](), [Google]() Tags: [2026](), [ai](), [ai-gateway](), [anthropic](), [cost](), [deepseek](), [google](), [models](), [open](), [openai](), [production](), [release](), [tokens]() ### AI overview The July 2026 AI Gateway Production Index reports that gateway token volume grew 29% month over month in June while spending grew 27% and average token prices remained flat. Open-weight models reached 29% of token volume for under 4% of spend, while Anthropic captured 61% of spending. The report also compares model and provider leadership across text, image, and video workloads. ### Source excerpt AI Gateway Production Index -- July 2026 Every month, AI Gateway routes tens of trillions of tokens between production applications and AI labs, giving us a view of what AI usage actually looks like in today's enterprise. We publish that view here. See the Production Index reports published in May and June. July 2026 summary The July index reports on AI Gateway data collected in June 2026. AI Gateway token volume grew 29% month over month and spend grew 27%. The price per token was flat after rising almost 20% in May. Open-weight models ran 29% of gateway tokens, up from 11% in April, on under 4% of spend. DeepSeek reached 22.6% of token volume, in third place and less than two points behind Google, and GLM 5.2 broke into the gateway's top models by volume two weeks after release. Anthropic took 61% of gateway spend on 32% of tokens, and captured 72% or more of spend in every high-stakes use case. Each modality has a different leader. OpenAI's GPT Image generated 53% of images, ahead of Google's Nano Banana at 39%. Chinese labs took two-thirds of video spend, while xAI's Grok Imagine generated 42% of videos, the most of any model. Claude Fable 5 was released June 9 and reached 22% of Opus 4.8's request volume in four days. A US export-control directive suspended access for the rest of the month. AI investment kept climbing, and the average token price went flat In June, token volume and spend grew at nearly the same rate, 29% and 27%, while the price per token remained steady. The month before, spend had grown more than twice as fast as volume, driving the average token cost up by almost 20%. The market spent more overall, but not more per token. Since April, open-weight models have climbed from a ninth of all token volume to nearly a third, at about a tenth of the average token price on the gateway. That alone should have pulled the price per token down. But as cheap volume rose in June, so too did closed-weight frontier prices, up about 12% per token. They offset e [Next page]()