# Runpod Blog.

Our team's insights on building better and scaling smarter.

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

## Announcing Runpod's Integration with SkyPilot

DevFeed: [Announcing Runpod's Integration with SkyPilot](<https://devfeed.tech/articles/announcing-runpod-s-integration-with-skypilot-140528.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-skypilot-integration>)

Author: Brendan McKeag

Published: 2026-11-17T22:44:11Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [skypilot](<https://devfeed.tech/topics/skypilot.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [AI Platform](<https://devfeed.tech/topics/ai-platform.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [batch](<https://devfeed.tech/tags/batch.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-orchestration](<https://devfeed.tech/tags/cloud-orchestration.md>), [flexible](<https://devfeed.tech/tags/flexible.md>), [integration](<https://devfeed.tech/tags/integration.md>), [jobs](<https://devfeed.tech/tags/jobs.md>), [llms](<https://devfeed.tech/tags/llms.md>), [multi-cloud](<https://devfeed.tech/tags/multi-cloud.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [skypilot](<https://devfeed.tech/tags/skypilot.md>)

### AI overview

The March 2024 announcement describes SkyPilot as a way to combine cloud providers and Kubernetes into a compute pool, automatically routing AI and batch workloads toward the cheapest available resources. Runpod's integration uses its CLI to provision on-demand pods and deploy serverless endpoints. The excerpt's setup begins with Runpod configuration and SkyPilot's Runpod package, then checks credentials with `sky check` and launches a cluster with `sky launch`. These setup instructions have changed: the source says current use requires `runpod>=1.6` and directs readers to SkyPilot's integration docs; its original command specifies `runpod>=1.5.1`.

### Source excerpt

Runpod now integrates with SkyPilot, enabling even more flexible scheduling and multi-cloud orchestration for LLMs, batch jobs, and custom AI workloads.

## Connect Google Colab to Runpod for Custom GPU Power

DevFeed: [Connect Google Colab to Runpod for Custom GPU Power](<https://devfeed.tech/articles/connect-google-colab-to-runpod-for-custom-gpu-power-140527.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/connect-google-colab-to-runpod-gpu>)

Author: Zhen Lu

Published: 2026-11-17T22:44:11Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [ssh](<https://devfeed.tech/topics/ssh.md>), [Google](<https://devfeed.tech/topics/google.md>)

Tags: [after](<https://devfeed.tech/tags/after.md>), [all](<https://devfeed.tech/tags/all.md>), [but](<https://devfeed.tech/tags/but.md>), [button](<https://devfeed.tech/tags/button.md>), [can](<https://devfeed.tech/tags/can.md>), [case](<https://devfeed.tech/tags/case.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connect](<https://devfeed.tech/tags/connect.md>), [custom](<https://devfeed.tech/tags/custom.md>), [google](<https://devfeed.tech/tags/google.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [guide](<https://devfeed.tech/tags/guide.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [interface](<https://devfeed.tech/tags/interface.md>), [more](<https://devfeed.tech/tags/more.md>), [need](<https://devfeed.tech/tags/need.md>), [port](<https://devfeed.tech/tags/port.md>), [power](<https://devfeed.tech/tags/power.md>), [shows](<https://devfeed.tech/tags/shows.md>), [this](<https://devfeed.tech/tags/this.md>)

### AI overview

The guide connects Google Colab to a Runpod GPU by forwarding the pod's Jupyter port through the user's machine. Start a Runpod instance with SSH Terminal Access enabled, set the Docker command to enable Jupyter's HTTP-over-websockets extension, and note the JUPYTER_PASSWORD. After the pod starts, append `-L 8888:localhost:8888` to its SSH connection command to forward port 8888 locally. In Colab, choose "Connect to a local runtime" and enter the local runtime address with the copied password after `?token=`. The author says this works with standard Runpod templates; custom images may need additional dependencies. The text omits the actual Docker command and full connection command, so those exact values are not available here.

### Source excerpt

Prefer Colab's interface but need more reliable compute? This guide shows you how to connect Google Colab to a Runpod instance via port forwarding.

## F5-TTS voice cloning: how to clone your voice with AI on Runpod

DevFeed: [F5-TTS voice cloning: how to clone your voice with AI on Runpod](<https://devfeed.tech/articles/f5-tts-voice-cloning-how-to-clone-your-voice-with-ai-on-runpod-131264.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/clone-your-voice-f5-tts>)

Author: Daniel Monedero-Tortola

Published: 2026-11-12T23:13:09Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Learning](<https://devfeed.tech/topics/learning.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [Whisper](<https://devfeed.tech/topics/whisper.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [checkpoint](<https://devfeed.tech/tags/checkpoint.md>), [clone](<https://devfeed.tech/tags/clone.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [f5-tts](<https://devfeed.tech/tags/f5-tts.md>), [generate](<https://devfeed.tech/tags/generate.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [license](<https://devfeed.tech/tags/license.md>), [model](<https://devfeed.tech/tags/model.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [pod](<https://devfeed.tech/tags/pod.md>), [recording](<https://devfeed.tech/tags/recording.md>), [recordings](<https://devfeed.tech/tags/recordings.md>), [selection](<https://devfeed.tech/tags/selection.md>), [speech](<https://devfeed.tech/tags/speech.md>), [tts](<https://devfeed.tech/tags/tts.md>), [tts-voice-cloning](<https://devfeed.tech/tags/tts-voice-cloning.md>), [voice](<https://devfeed.tech/tags/voice.md>), [voice-cloning](<https://devfeed.tech/tags/voice-cloning.md>)

### AI overview

The tutorial explains how to fine-tune F5-TTS on at least 30 minutes of your own speech using a Runpod GPU pod, then generate speech from text in your voice. It covers recording matched audio and transcripts, preparing 24 kHz clips, training and comparing checkpoints, choosing reference audio, and downloading the model. It recommends a GPU with at least 24 GB VRAM and a persistent 50 GB network volume; training 30 minutes of audio for 200 epochs reportedly took about 75 minutes on an L4. Listen to samples and use an earlier checkpoint if later output becomes metallic from overtraining. The F5-TTS code is MIT-licensed, but its pretrained weights and fine-tuned derivatives are non-commercial; the tutorial suggests Chatterbox for commercial use. Keep the checkpoint private, since it can produce audio resembling your voice.

### Source excerpt

Fine-tune the open-source F5-TTS model on your own recordings and generate speech in your voice on a Runpod GPU pod. Covers recording, dataset prep, training, checkpoint selection, and the F5-TTS license.

## AMD MI300X vs NVIDIA H100 SXM: Mixtral 8x7B Inference

DevFeed: [AMD MI300X vs NVIDIA H100 SXM: Mixtral 8x7B Inference](<https://devfeed.tech/articles/amd-mi300x-vs-nvidia-h100-sxm-mixtral-8x7b-inference-129988.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/amd-mi300x-vs-nvidia-h100-sxm-performance-comparison>)

Author: Marut Pandya

Published: 2026-11-12T00:09:51Z

Content type: comparison

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [mixtral](<https://devfeed.tech/topics/mixtral.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>), [CUDA](<https://devfeed.tech/topics/cuda.md>), [rocm](<https://devfeed.tech/topics/rocm.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [ai-training-and-inference](<https://devfeed.tech/tags/ai-training-and-inference.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [amd](<https://devfeed.tech/tags/amd.md>), [applications](<https://devfeed.tech/tags/applications.md>), [batch](<https://devfeed.tech/tags/batch.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cost](<https://devfeed.tech/tags/cost.md>), [h100](<https://devfeed.tech/tags/h100.md>), [inference](<https://devfeed.tech/tags/inference.md>), [mistral](<https://devfeed.tech/tags/mistral.md>), [mixtral](<https://devfeed.tech/tags/mixtral.md>), [model](<https://devfeed.tech/tags/model.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-h100](<https://devfeed.tech/tags/nvidia-h100.md>), [performance](<https://devfeed.tech/tags/performance.md>), [results](<https://devfeed.tech/tags/results.md>), [vs](<https://devfeed.tech/tags/vs.md>)

### AI overview

In Runpod's June 2024 Mixtral 8x7B inference benchmark, the MI300X's advantage depended on batch size: it was more cost effective at batches 1-4 and 256 or higher, while the H100 SXM had stronger throughput at small to medium sizes and MI300X results were weaker in the middle range. The comparison used vLLM batched offline inference in FP16 with 128-token inputs and outputs. Mixtral required two H100 SXMs because it exceeded one H100's 80 GB of VRAM; one 192 GB MI300X could hold it. To compare two-GPU configurations, MI300X throughput was estimated by doubling the preceding batch-size result, so those figures are extrapolated. The article also reports lower MI300X latency and more consistent performance under higher serving loads. Pricing used Runpod Secure Cloud rates at the time, and the article cautions that those rates and derived token costs are historical snapshots.

### Source excerpt

Runpod benchmarks AMD's MI300X against Nvidia's H100 SXM using Mistral's Mixtral 8x7B model. The results highlight performance and cost trade-offs across.

## How Do I Transfer Data Into My Runpod?

DevFeed: [How Do I Transfer Data Into My Runpod?](<https://devfeed.tech/articles/how-do-i-transfer-data-into-my-runpod-127619.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/transfer-data-into-runpod>)

Author: Brendan McKeag

Published: 2026-11-10T22:57:50Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Data transmission](<https://devfeed.tech/topics/data-transmission.md>), [rsync](<https://devfeed.tech/topics/rsync.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [about](<https://devfeed.tech/tags/about.md>), [account](<https://devfeed.tech/tags/account.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [agents](<https://devfeed.tech/tags/agents.md>), [api](<https://devfeed.tech/tags/api.md>), [api-key](<https://devfeed.tech/tags/api-key.md>), [arguments](<https://devfeed.tech/tags/arguments.md>), [based](<https://devfeed.tech/tags/based.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [files](<https://devfeed.tech/tags/files.md>), [guide](<https://devfeed.tech/tags/guide.md>), [large](<https://devfeed.tech/tags/large.md>), [large-datasets](<https://devfeed.tech/tags/large-datasets.md>), [local](<https://devfeed.tech/tags/local.md>), [need](<https://devfeed.tech/tags/need.md>), [pod](<https://devfeed.tech/tags/pod.md>), [this](<https://devfeed.tech/tags/this.md>)

### AI overview

For large datasets, the guide recommends hosting files online and pulling them into a pod, since uploads from a local computer are limited by its internet connection. Options include Runpod's runpodctl for transfers between a local machine and pods or directly between pods; it needs terminal access but no HTTP or TCP ports. You can also drag files into supported browser editors, use wget for internet-hosted files (some sites require an API key or special downloader), or clone Git and Hugging Face repositories; install and initialize Git LFS for files over 100 MB. SCP and rsync are additional options that use the pod's external IP and TCP port. The excerpt says Runpod charges no ingress or egress fees; storage charges depend on where data lands.

### Source excerpt

Need to move files into your Runpod? This guide explains the fastest, most reliable ways to transfer large datasets into your pod, whether local or.

## What's new in Runpod for Agents: November 2026

DevFeed: [What's new in Runpod for Agents: November 2026](<https://devfeed.tech/articles/what-s-new-in-runpod-for-agents-november-2026-127094.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-agent-skills-mcp-server>)

Author: Bohdan Yurov

Published: 2026-11-10T15:32:46Z

Content type: article

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [hardened ai agent skills](<https://devfeed.tech/topics/hardened-ai-agent-skills.md>), [context-architecture](<https://devfeed.tech/topics/context-architecture.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>), [MSP MCP](<https://devfeed.tech/topics/msp-mcp.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [agents](<https://devfeed.tech/tags/agents.md>), [api](<https://devfeed.tech/tags/api.md>), [ci](<https://devfeed.tech/tags/ci.md>), [cli](<https://devfeed.tech/tags/cli.md>), [cold-start](<https://devfeed.tech/tags/cold-start.md>), [comfyui](<https://devfeed.tech/tags/comfyui.md>), [credentials](<https://devfeed.tech/tags/credentials.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-centers](<https://devfeed.tech/tags/data-centers.md>), [dc](<https://devfeed.tech/tags/dc.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [evals](<https://devfeed.tech/tags/evals.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-server](<https://devfeed.tech/tags/mcp-server.md>), [november-2026](<https://devfeed.tech/tags/november-2026.md>), [rules](<https://devfeed.tech/tags/rules.md>), [skills](<https://devfeed.tech/tags/skills.md>), [tests](<https://devfeed.tech/tags/tests.md>), [what-s-new](<https://devfeed.tech/tags/what-s-new.md>)

### AI overview

Runpod's agent setup pairs a hosted MCP server, which exposes typed infrastructure operations, with Markdown skills that guide tool choice and procedures. The article's central finding is that skills go stale unless their claims are sourced and tested. Runpod describes a concept graph whose rules cite documentation, API references or probes, plus build checks, more than 80 skill eval cases, and 52 live-account user stories run with and without skills. Live recipe runs and agent reports feed fixes back into the skills, rules and tests. One example rule delays creating a network volume until stock selection determines the Pod's data center, avoiding repeated volume deletion when trying another location. The article says this system helps agents diagnose issues from worker logs and follow correct volume and deployment procedures; it does not provide measured performance results.

### Source excerpt

Skills tell an agent what to do and the MCP server does it. See how sourced rules, evals and live-account tests keep Runpod's agent skills accurate.

## Runpod Terraform provider 2.0: migrate from 1.x

DevFeed: [Runpod Terraform provider 2.0: migrate from 1.x](<https://devfeed.tech/articles/runpod-terraform-provider-2-0-migrate-from-1-x-124883.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-terraform-provider-2-0-migration>)

Author: The Team at Runpod

Published: 2026-11-10T06:10:33Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Terraform](<https://devfeed.tech/topics/terraform.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [api-key](<https://devfeed.tech/tags/api-key.md>), [argument](<https://devfeed.tech/tags/argument.md>), [auth](<https://devfeed.tech/tags/auth.md>), [backends](<https://devfeed.tech/tags/backends.md>), [base](<https://devfeed.tech/tags/base.md>), [ci](<https://devfeed.tech/tags/ci.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agent](<https://devfeed.tech/tags/coding-agent.md>), [commit](<https://devfeed.tech/tags/commit.md>), [configuration](<https://devfeed.tech/tags/configuration.md>), [migrate](<https://devfeed.tech/tags/migrate.md>), [plan](<https://devfeed.tech/tags/plan.md>), [pod](<https://devfeed.tech/tags/pod.md>), [rest](<https://devfeed.tech/tags/rest.md>), [terraform](<https://devfeed.tech/tags/terraform.md>), [terraform-provider](<https://devfeed.tech/tags/terraform-provider.md>)

### AI overview

Runpod Terraform provider 2.0 is generated from the REST v2 API, giving its attributes documented v2 mappings; version 1.x used REST v1, which Runpod is retiring. The provider manages Pods, Serverless endpoints, templates, volumes, Instant Clusters, and registry credentials through Terraform. It requires Terraform 1.6 or later, and the article recommends supplying the API key through `RUNPOD_API_KEY`. Plans let you review changes before applying billable resources; running Pods bill by the minute, so destroy them when finished or set `desired_state = "EXITED"` to stop them while retaining configuration. Migration is breaking: 2.0.0 has no state upgraders, so resources need configuration edits and state actions. Some 1.x Pods may be replaced because v2 reads omit eight create-time attributes; replacement permanently deletes the Pod and anything not on a network volume. The guide advises reviewing planned replacements and describes leaving Pods on 1.x or adopting them with `ignore_changes` before the first apply.

### Source excerpt

## How Runpod's agent skills and MCP server work together An API tells an agent what it can do, not whether it should. See how skills, a concept graph and live-account tests keep agents right on Runpod. Blog Put Runpod in the same Terraform plan as the rest of your stack. Deploy your first Pod, let an agent write the plan and migrate from 1.x. Terraform lets you describe infrastructure as code, preview the changes and apply them.

## Connect Claude Code, Codex or Cursor to Runpod with MCP

DevFeed: [Connect Claude Code, Codex or Cursor to Runpod with MCP](<https://devfeed.tech/articles/connect-claude-code-codex-or-cursor-to-runpod-with-mcp-124885.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/set-up-runpod-coding-agent-mcp>)

Author: The Team at Runpod

Published: 2026-11-10T06:09:50Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [codex](<https://devfeed.tech/topics/codex.md>), [cursor](<https://devfeed.tech/topics/cursor.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [ChatGPT](<https://devfeed.tech/topics/chatgpt.md>)

Tags: [account](<https://devfeed.tech/tags/account.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [agents](<https://devfeed.tech/tags/agents.md>), [api](<https://devfeed.tech/tags/api.md>), [blog](<https://devfeed.tech/tags/blog.md>), [can](<https://devfeed.tech/tags/can.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [ci](<https://devfeed.tech/tags/ci.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-code](<https://devfeed.tech/tags/claude-code.md>), [cleanup](<https://devfeed.tech/tags/cleanup.md>), [code](<https://devfeed.tech/tags/code.md>), [codex](<https://devfeed.tech/tags/codex.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agent](<https://devfeed.tech/tags/coding-agent.md>), [connect](<https://devfeed.tech/tags/connect.md>), [cursor](<https://devfeed.tech/tags/cursor.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

Runpod connects coding agents through tested skills for deployment procedures and a hosted MCP server for tools and current documentation. The excerpt says Claude Code and Codex install a plugin containing both; ChatGPT uses a plugin listing, while Cursor and most other clients use two commands. Setup registers https://mcp.runpod.io/ and requires sign-in, by browser or an API key when browser authentication is unavailable. MCP tools carry read-only or destructive labels so users can control permitted actions; in Claude Code CI runs, destructive tools remain blocked unless explicitly allowed. An empty Pod list can indicate a successful connection, but if Pods were expected, the agent may be signed into a different account. Register only one server URL because the legacy mcp.getrunpod.io address has separate sign-in.

### Source excerpt

## How Runpod's agent skills and MCP server work together An API tells an agent what it can do, not whether it should. See how skills, a concept graph and live-account tests keep agents right on Runpod. Blog One prompt hands Runpod deploys and cleanup to your coding agent. See install steps for Claude Code, Codex, ChatGPT, Cursor and more, plus a CI path. You can hand Runpod setup, deploys and cleanup to your coding agent with one prompt.

## What's new in Runpod for Agents: November 2026

DevFeed: [What's new in Runpod for Agents: November 2026](<https://devfeed.tech/articles/what-s-new-in-runpod-for-agents-november-2026-124887.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/whats-new-runpod-agents-november-2026>)

Author: The Team at Runpod

Published: 2026-11-10T06:09:34Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [Documentation](<https://devfeed.tech/topics/documentation.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [self-service](<https://devfeed.tech/topics/self-service.md>)

Tags: [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [api](<https://devfeed.tech/tags/api.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [claude-code](<https://devfeed.tech/tags/claude-code.md>), [codex](<https://devfeed.tech/tags/codex.md>), [console](<https://devfeed.tech/tags/console.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cursor](<https://devfeed.tech/tags/cursor.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [diagnose](<https://devfeed.tech/tags/diagnose.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-server](<https://devfeed.tech/tags/mcp-server.md>), [november-2026](<https://devfeed.tech/tags/november-2026.md>), [pods](<https://devfeed.tech/tags/pods.md>), [what-s-new](<https://devfeed.tech/tags/what-s-new.md>)

### AI overview

Runpod for Agents pairs current documentation and tested procedures with API access so coding agents can handle infrastructure tasks. A setup prompt installs the skills and MCP server in Claude Code, Codex, Cursor and other supported clients; sign-in uses a browser or an API key if browser sign-in is unavailable. The MCP server adds a knowledge base and tools for searching concepts, guides and documentation, with sources returned for documentation searches. Ten skills cover common deployment, troubleshooting, operations, cost and migration tasks; the plugin also includes 25 worked reference tasks, most tested on live infrastructure. The Runpod plugin works in Claude Code and ChatGPT. Terraform provider 2.0 and a 0.x TypeScript SDK are also generated from REST v2, aligning scripts and agent tools around the same API. The text notes that skills prefer Secure Cloud and set a minimum CUDA version.

### Source excerpt

## How Runpod's agent skills and MCP server work together An API tells an agent what it can do, not whether it should. See how skills, a concept graph and live-account tests keep agents right on Runpod. Blog Let your coding agent deploy Pods, diagnose failures and shut down what you no longer need, while you decide which actions it takes without asking. Runpod for Agents lets you hand infrastructure work to your coding agent.

## Build container images on your Runpod Pod with Runpod Weld

DevFeed: [Build container images on your Runpod Pod with Runpod Weld](<https://devfeed.tech/articles/build-container-images-on-your-runpod-pod-with-runpod-weld-124884.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-weld-build-images-on-your-pod>)

Author: Roman Przybyla

Published: 2026-11-09T23:41:48Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Container Image Building](<https://devfeed.tech/topics/container-image-building.md>), [container images](<https://devfeed.tech/topics/container-images.md>), [Dockerfile](<https://devfeed.tech/topics/dockerfile.md>), [Docker Hub](<https://devfeed.tech/topics/docker-hub.md>), [Docker Image](<https://devfeed.tech/topics/docker-image.md>)

Tags: [build](<https://devfeed.tech/tags/build.md>), [container](<https://devfeed.tech/tags/container.md>), [container-images](<https://devfeed.tech/tags/container-images.md>), [demo](<https://devfeed.tech/tags/demo.md>), [docker](<https://devfeed.tech/tags/docker.md>), [image](<https://devfeed.tech/tags/image.md>), [images](<https://devfeed.tech/tags/images.md>), [inside](<https://devfeed.tech/tags/inside.md>), [pod](<https://devfeed.tech/tags/pod.md>)

### AI overview

Runpod Weld builds and pushes container images from a Runpod Pod without a Docker daemon; a separate container runtime is needed to run them. The guide shows checking the Pod with `rpweld doctor`, installing Weld on Linux amd64 (Debian 12 is the validated target), preparing a Dockerfile, and building and pushing an image to Docker Hub. Weld caches build stages, but builds are slower than local Docker, support only a subset of Dockerfile syntax, and are alpha software. Builds need Pod capabilities and filesystem permissions for `RUN` steps; Weld is not a security sandbox, and trusted Dockerfiles are recommended because `RUN` commands can access the Pod. Its image store and cache are separate from Docker's, finished builds are not automatically deleted, and registry tokens are base64-encoded rather than encrypted. The guide demonstrates deploying the pushed image on Runpod or running it on a Linux amd64 Docker host, and notes that private images require registry credentials.

### Source excerpt

Runpod Weld builds and pushes container images from inside a Runpod Pod, no Docker daemon needed. Build a demo image, push it, and run it on a Pod.

## How to get started with Qwen3.8-Flash-Next on Runpod Serverless

DevFeed: [How to get started with Qwen3.8-Flash-Next on Runpod Serverless](<https://devfeed.tech/articles/how-to-get-started-with-qwen3-8-flash-next-on-runpod-serverless-124880.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/getting-started-with-qwen3-8-flash-next-on-runpod-serverless>)

Author: Chris Vela

Published: 2026-11-09T23:35:25Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Serverless](<https://devfeed.tech/topics/serverless.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [CUDA Graphs](<https://devfeed.tech/topics/cuda-graphs.md>), [huggingface\_hub](<https://devfeed.tech/topics/huggingface-hub.md>)

Tags: [deploy](<https://devfeed.tech/tags/deploy.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [qwen3-8-flash-next](<https://devfeed.tech/tags/qwen3-8-flash-next.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [vllm](<https://devfeed.tech/tags/vllm.md>)

### AI overview

The Runpod Hub vLLM worker template described in the excerpt uses vLLM 0.28.0, below Qwen3.8-Flash-Next's stated 0.29 minimum, so the validated deployment uses vLLM's `vllm/vllm-openai:qwen38-flash-next` image. The authors ran the FP8 model on 4x H200 with tensor parallelism and expert parallelism; Hopper requires `--enable-expert-parallel --moe-backend triton`. Set `NCCL_NVLS_ENABLE=0` to avoid NVLS initialization failures. `--enforce-eager` shortens startup by skipping compilation and CUDA graph capture, but makes early requests much slower and steady-state decode slower than the compiled setup. The reported warm-host startup was about 11 minutes; a fully cold model download may take 20-25 minutes. The excerpt recommends load-balancing Serverless for bursty API traffic and pods for interactive work or latency-sensitive use. It reports successful checks for OpenAI-compatible requests, streaming, tool calling, instruct mode, and vision.

### Source excerpt

Deploy Qwen3.8-Flash-Next on Runpod with vLLM's dedicated image. Validated flags, 4x H200 hardware math, measured cold starts, and the NCCL fix you need on day one.

## Introducing A40 GPUs for Machine Learning | Runpod

DevFeed: [Introducing A40 GPUs for Machine Learning | Runpod](<https://devfeed.tech/articles/introducing-a40-gpus-for-machine-learning-runpod-124881.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/introducing-a40-gpus-machine-learning>)

Author: Brendan McKeag

Published: 2026-11-09T23:35:25Z

Content type: article

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [availability](<https://devfeed.tech/tags/availability.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [cost](<https://devfeed.tech/tags/cost.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [high-performance](<https://devfeed.tech/tags/high-performance.md>), [llms](<https://devfeed.tech/tags/llms.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [value](<https://devfeed.tech/tags/value.md>)

### AI overview

Runpod's A40 offers 48 GB of VRAM for machine learning workloads. The excerpt says users can choose A40 when deploying a Pod, or select a 48 GB GPU instance and then A40 as the GPU type in Serverless. It refers to benchmarks comparing A40 with H100 and a table of model configurations and prices per million tokens, but does not include their results, so comparative performance or cost conclusions cannot be drawn from this text.

### Source excerpt

NVIDIA A40 GPUs on Runpod pair high performance with low cost and strong availability, making them a good value for fine-tuning LLMs and machine learning.

## Run DeepSeek R1 on Just 480GB of VRAM

DevFeed: [Run DeepSeek R1 on Just 480GB of VRAM](<https://devfeed.tech/articles/run-deepseek-r1-on-just-480gb-of-vram-124882.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/run-deepseek-r1-low-vram>)

Author: Brendan McKeag

Published: 2026-11-09T23:35:25Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

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

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [challenge](<https://devfeed.tech/tags/challenge.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cutlass](<https://devfeed.tech/tags/cutlass.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [deepseek-r1](<https://devfeed.tech/tags/deepseek-r1.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [download](<https://devfeed.tech/tags/download.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [performance](<https://devfeed.tech/tags/performance.md>)

### AI overview

The post describes running DeepSeek R1's Q4 4-bit quantization through a Runpod KoboldCPP template, requiring at least 480GB of VRAM. The suggested setup is the "KoboldCpp - DeepSeek-R1 - 6x80GB of VRAM" template, which provides an OpenAI-compatible endpoint on port 5001; download and loading are estimated at about 20 minutes. Listed Secure Cloud costs are $10.74/hour on 6xA100s or $21.16/hour on 4xH200s. The post also says you can lower KoboldCPP's `--gpulayers` setting to offload layers to system RAM and use fewer GPUs, at the cost of a sizable performance hit. It cautions that community templates may have bugs and mentions an official KoboldCPP package as an alternative requiring manual download configuration.

### Source excerpt

DeepSeek R1 remains one of the top open-source models. This post shows how to run it efficiently on just 480GB of VRAM without sacrificing performance.

## How a Solo Dev Built an AI for Dads, No GPU, No Team, Just $5

DevFeed: [How a Solo Dev Built an AI for Dads, No GPU, No Team, Just $5](<https://devfeed.tech/articles/how-a-solo-dev-built-an-ai-for-dads-no-gpu-no-team-just-5-124886.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/solo-dev-ai-for-dads-runpod>)

Author: Alyssa Mazzina

Published: 2026-11-09T23:35:25Z

Content type: news

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [Data Quality](<https://devfeed.tech/topics/data-quality.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [accessible](<https://devfeed.tech/tags/accessible.md>), [account](<https://devfeed.tech/tags/account.md>), [adapters](<https://devfeed.tech/tags/adapters.md>), [advice](<https://devfeed.tech/tags/advice.md>), [ai](<https://devfeed.tech/tags/ai.md>), [build](<https://devfeed.tech/tags/build.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [developer](<https://devfeed.tech/tags/developer.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [just](<https://devfeed.tech/tags/just.md>), [one](<https://devfeed.tech/tags/one.md>), [product](<https://devfeed.tech/tags/product.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [solo](<https://devfeed.tech/tags/solo.md>), [solo-developer](<https://devfeed.tech/tags/solo-developer.md>), [team](<https://devfeed.tech/tags/team.md>), [this](<https://devfeed.tech/tags/this.md>), [used](<https://devfeed.tech/tags/used.md>), [writing](<https://devfeed.tech/tags/writing.md>)

### AI overview

Benoit Rossignol built DadAI, an emotional support assistant for new fathers, by fine-tuning Mistral 7B with QLoRA and PEFT on a Runpod RTX 4090. The article reports three training epochs in under six minutes, a final loss of 2.23, and a total cost under $5; he kept idle costs low by stopping the pod between sessions. Deployment hit a format mismatch: his GPTQ-quantized model and LoRA adapters were incompatible with LocalAI's llama.cpp backend, which expects GGUF. Attempts to build LocalAI on a Mac M1 and Runpod failed, leaving the current model runnable from the command line. Rossignol says he is working on a Hugging Face Space or simple UI demo. DadAI is presented as a judgment-free companion, not a replacement for human support.

### Source excerpt

No GPU. No team. Just $5. This is how one solo developer used Runpod Serverless to build and deploy a working AI product, "AI for Dads", without writing.

## Flash, now the Nix way: run the Flash CLI straight from GitHub

DevFeed: [Flash, now the Nix way: run the Flash CLI straight from GitHub](<https://devfeed.tech/articles/flash-now-the-nix-way-run-the-flash-cli-straight-from-github-121789.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/flash-nix>)

Author: Dave Seddon

Published: 2026-11-06T15:08:43Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Nix](<https://devfeed.tech/topics/nix.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [cli](<https://devfeed.tech/tags/cli.md>), [flash](<https://devfeed.tech/tags/flash.md>), [github](<https://devfeed.tech/tags/github.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [nix](<https://devfeed.tech/tags/nix.md>), [serverless](<https://devfeed.tech/tags/serverless.md>)

### AI overview

Flash's Nix flake lets developers run the CLI directly from GitHub with `nix run github:runpod/flash`, install it on their PATH, or enter a development shell. Its `flake.lock` pins Nix inputs for a reproducible local environment; Flash separately packages and deploys workloads to Runpod GPUs, so local CUDA and GPU drivers are unnecessary. The excerpt also shows a project using Python 3.14 locally and Python 3.12 on the worker, and notes that `pip install runpod-flash` remains available. Worker package version pins alone do not lock transitive dependencies, so the example also pins the model revision. The flake supports Linux x86-64 and ARM64 and Apple Silicon; Windows users can work through WSL2.

### Source excerpt

## Building Runpod templates: Your first custom template Build a custom Docker image with PyTorch and a short GPU-check script, push it to Docker Hub, and save it as a private Runpod Pod template so every new Pod starts with your setup in place. Blog Flash now ships a first-class Nix flake. Run the CLI with nix run github:runpod/flash, get a pinned dev shell, and send GPU work to Runpod Serverless.

## Global Volumes aren't "network volumes, but better." Here's what they're for.

DevFeed: [Global Volumes aren't "network volumes, but better." Here's what they're for.](<https://devfeed.tech/articles/global-volumes-aren-t-network-volumes-but-better-here-s-what-they-re-for-121790.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/global-volumes-arent-network-volumes>)

Author: Brendan McKeag

Published: 2026-11-06T15:08:43Z

Content type: article

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [data lake](<https://devfeed.tech/topics/data-lake.md>), [self-service](<https://devfeed.tech/topics/self-service.md>)

Tags: [blog](<https://devfeed.tech/tags/blog.md>), [build](<https://devfeed.tech/tags/build.md>), [building](<https://devfeed.tech/tags/building.md>), [but](<https://devfeed.tech/tags/but.md>), [changes](<https://devfeed.tech/tags/changes.md>), [custom](<https://devfeed.tech/tags/custom.md>), [docker](<https://devfeed.tech/tags/docker.md>), [docker-hub](<https://devfeed.tech/tags/docker-hub.md>), [docker-image](<https://devfeed.tech/tags/docker-image.md>), [first](<https://devfeed.tech/tags/first.md>), [global](<https://devfeed.tech/tags/global.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hub](<https://devfeed.tech/tags/hub.md>), [image](<https://devfeed.tech/tags/image.md>), [like](<https://devfeed.tech/tags/like.md>), [mount](<https://devfeed.tech/tags/mount.md>), [network](<https://devfeed.tech/tags/network.md>), [object](<https://devfeed.tech/tags/object.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [volumes](<https://devfeed.tech/tags/volumes.md>)

### AI overview

Runpod Global Volumes let Pods in different datacenters access shared data, but they use object storage rather than a filesystem. They work well for large files that are written once and read whole, such as models and LoRAs; they can be slow or unreliable for software installs, in-place edits, locks, symlinks, Git operations, and many small files. For ComfyUI, keep the install and custom nodes on local or network storage near the GPU, and keep models on the Global Volume. Custom nodes can be stored there as an archive and unpacked onto local storage at startup. This lets model data remain shared while Pods can use available GPUs across datacenters.

### Source excerpt

## Building Runpod templates: Your first custom template Build a custom Docker image with PyTorch and a short GPU-check script, push it to Docker Hub, and save it as a private Runpod Pod template so every new Pod starts with your setup in place. Blog Global Volumes mount like a network volume but run on object storage. That changes what you should put on them.

## Introducing the Runpod Hub: Deploy Open Source AI Repos

DevFeed: [Introducing the Runpod Hub: Deploy Open Source AI Repos](<https://devfeed.tech/articles/introducing-the-runpod-hub-deploy-open-source-ai-repos-121792.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-hub-launch-open-source-ai-repos>)

Author: Alyssa Mazzina

Published: 2026-11-06T14:47:06Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-developers](<https://devfeed.tech/tags/ai-developers.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [applications](<https://devfeed.tech/tags/applications.md>), [audio](<https://devfeed.tech/tags/audio.md>), [comfyui](<https://devfeed.tech/tags/comfyui.md>), [community](<https://devfeed.tech/tags/community.md>), [contributed](<https://devfeed.tech/tags/contributed.md>), [contributions](<https://devfeed.tech/tags/contributions.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [fork](<https://devfeed.tech/tags/fork.md>), [hub](<https://devfeed.tech/tags/hub.md>), [image](<https://devfeed.tech/tags/image.md>), [llms](<https://devfeed.tech/tags/llms.md>), [marketplace](<https://devfeed.tech/tags/marketplace.md>), [models](<https://devfeed.tech/tags/models.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

Runpod Hub offers preconfigured, community-contributed open-source AI repositories that users can browse, fork, and deploy on Runpod's serverless infrastructure. The excerpt says contributors can connect a GitHub repo and provide a Dockerfile; Runpod handles containerization and provisioning, producing a scalable endpoint. Examples include ComfyUI for visual image-generation workflows, Whisper for speech recognition, and Mochi 1 for video generation, alongside community contributions Axolotl for fine-tuning and Ollama for local LLM capabilities. The excerpt presents the Hub as a way to avoid manual container setup and deployment scripting, and says listings are open source and GitHub-connected.

### Source excerpt

The Runpod Hub is here, a creator-powered marketplace for open source AI. Browse, fork, and deploy prebuilt repos for LLMs, image models, video.

## Deep Dive Into Creating and Listing on the Runpod Hub

DevFeed: [Deep Dive Into Creating and Listing on the Runpod Hub](<https://devfeed.tech/articles/deep-dive-into-creating-and-listing-on-the-runpod-hub-121788.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/deep-dive-runpod-hub>)

Author: Brendan McKeag

Published: 2026-11-06T14:47:06Z

Content type: article

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Serverless](<https://devfeed.tech/topics/serverless.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Container Image Building](<https://devfeed.tech/topics/container-image-building.md>), [project release](<https://devfeed.tech/topics/project-release.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-deployment](<https://devfeed.tech/tags/ai-deployment.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [deep-dive](<https://devfeed.tech/tags/deep-dive.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [github](<https://devfeed.tech/tags/github.md>), [hub](<https://devfeed.tech/tags/hub.md>), [json](<https://devfeed.tech/tags/json.md>), [model](<https://devfeed.tech/tags/model.md>), [native](<https://devfeed.tech/tags/native.md>), [release](<https://devfeed.tech/tags/release.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [technical](<https://devfeed.tech/tags/technical.md>)

### AI overview

Runpod Hub uses GitHub releases to trigger builds and index serverless deployments: webhooks monitor a connected repository, and a release starts validation and infrastructure creation. A listing requires handler.py, a Dockerfile, hub.json and tests.json; the configuration files can live at the repository root or in .runpod, which takes precedence. The tests specification defines request cases, timeouts, GPU type and count, test environment variables, and supported CUDA versions. To submit, connect a repository, add the required files, create a release, and wait for Runpod staff review; approved listings can then be deployed, with initial endpoint settings derived from the JSON files. Listings are open source and GitHub connected, enabling users to inspect, fork, and modify implementations. The excerpt says creating a Hub listing is nearly identical to creating a serverless endpoint, but Hub offerings require staff vetting.

### Source excerpt

A deep technical dive into how the Runpod Hub streamlines serverless AI deployment with a GitHub-native, release-triggered model. Learn how hub.json and.

## Building Runpod templates: Your first custom template

DevFeed: [Building Runpod templates: Your first custom template](<https://devfeed.tech/articles/building-runpod-templates-your-first-custom-template-119778.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/building-runpod-templates-first-custom-template>)

Author: Roman Przybyla

Published: 2026-11-06T00:59:56Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Container Image Building](<https://devfeed.tech/topics/container-image-building.md>), [Docker Hub](<https://devfeed.tech/topics/docker-hub.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>), [Snippet](<https://devfeed.tech/topics/snippet.md>)

Tags: [build](<https://devfeed.tech/tags/build.md>), [custom](<https://devfeed.tech/tags/custom.md>), [docker](<https://devfeed.tech/tags/docker.md>), [docker-hub](<https://devfeed.tech/tags/docker-hub.md>), [docker-image](<https://devfeed.tech/tags/docker-image.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [guide](<https://devfeed.tech/tags/guide.md>), [pod](<https://devfeed.tech/tags/pod.md>), [pytorch](<https://devfeed.tech/tags/pytorch.md>), [template](<https://devfeed.tech/tags/template.md>), [templates](<https://devfeed.tech/tags/templates.md>), [test](<https://devfeed.tech/tags/test.md>)

### AI overview

The guide builds a reusable Runpod Pod setup by baking PyTorch and a GPU check script into a Docker image, publishing it to Docker Hub, and saving its settings in a private Runpod template. It pins the base image by digest, targets linux/amd64, and tests the script locally on CPU before checking normal container startup and running a GPU calculation in a Pod. Reusable software must be added to the Dockerfile and rebuilt; changes made in a running Pod are not captured by the template. A private template does not make a public image private, while private registries require pull credentials configured in Runpod. The guide also stresses matching the image to GPU and host CUDA support, reviewing deployment costs, and accounting for temporary container storage, persistent volume billing, and data loss when resources are terminated.

### Source excerpt

Step-by-step guide to building a custom PyTorch Docker image, pushing it to Docker Hub, and saving it as a reusable Runpod Pod template with a GPU test.

## How to use Kimi K3.1 Pro and K3.1 Pro Highspeed on Runpod

DevFeed: [How to use Kimi K3.1 Pro and K3.1 Pro Highspeed on Runpod](<https://devfeed.tech/articles/how-to-use-kimi-k3-1-pro-and-k3-1-pro-highspeed-on-runpod-117937.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/kimi-k3-1>)

Author: Jessica Garson Beauchemin

Published: 2026-11-04T19:50:18Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Documentation](<https://devfeed.tech/topics/documentation.md>), [cURL](<https://devfeed.tech/topics/curl.md>), [OpenAI](<https://devfeed.tech/topics/openai.md>)

Tags: [curl](<https://devfeed.tech/tags/curl.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [kimi](<https://devfeed.tech/tags/kimi.md>), [kimi-k3-1-pro](<https://devfeed.tech/tags/kimi-k3-1-pro.md>), [openai](<https://devfeed.tech/tags/openai.md>), [playground](<https://devfeed.tech/tags/playground.md>), [public](<https://devfeed.tech/tags/public.md>), [python](<https://devfeed.tech/tags/python.md>), [python-sdk](<https://devfeed.tech/tags/python-sdk.md>), [sdk](<https://devfeed.tech/tags/sdk.md>)

### AI overview

Kimi K3.1 Pro and K3.1 Pro Highspeed are available through Runpod Public Endpoints. The article describes three ways to try them: select a model and adjust prompt and generation settings in the Runpod Playground; send a curl request using a Runpod API key and poll the job every two seconds; or call the OpenAI-compatible endpoint with the OpenAI Python client. The curl and Python examples let callers choose either model by name. The Python route requires Python 3.9 or higher and the OpenAI client; the documented setup also requires outbound HTTPS access to api.runpod.ai, and curl examples use jq to format or filter JSON. The article describes Highspeed as two to three times faster than Pro, while Pro is positioned for long-horizon tasks and has a one-million-token context window.

### Source excerpt

Kimi K3.1 Pro and K3.1 Pro Highspeed are live on Runpod Public Endpoints. Get started in the Playground, with curl, or with the OpenAI Python SDK.

## Runpod Achieves SOC 2 Type II Certification

DevFeed: [Runpod Achieves SOC 2 Type II Certification](<https://devfeed.tech/articles/runpod-achieves-soc-2-type-ii-certification-112994.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/runpod-achieves-soc-2-type-ii-certification>)

Author: Brendan McKeag

Published: 2026-11-02T22:49:41Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [soc 2](<https://devfeed.tech/topics/soc-2.md>)

Tags: [access-controls](<https://devfeed.tech/tags/access-controls.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [audit](<https://devfeed.tech/tags/audit.md>), [audits](<https://devfeed.tech/tags/audits.md>), [availability](<https://devfeed.tech/tags/availability.md>), [certification](<https://devfeed.tech/tags/certification.md>), [change](<https://devfeed.tech/tags/change.md>), [change-management](<https://devfeed.tech/tags/change-management.md>), [changes](<https://devfeed.tech/tags/changes.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [code](<https://devfeed.tech/tags/code.md>), [company](<https://devfeed.tech/tags/company.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [security](<https://devfeed.tech/tags/security.md>), [security-controls](<https://devfeed.tech/tags/security-controls.md>), [soc](<https://devfeed.tech/tags/soc.md>), [soc-2](<https://devfeed.tech/tags/soc-2.md>), [standards](<https://devfeed.tech/tags/standards.md>)

### AI overview

Runpod says it achieved SOC 2 Type II certification after a six-month observation period, demonstrating that its controls operated consistently over time; Type I had assessed their design at a point in time. The audit examined areas including access controls, infrastructure security, incident response, change management, and vendor oversight. Runpod says its report is available through Drata after a signed NDA. Its certification covers Runpod's own information security management system; the third-party data centers underlying Secure Cloud hold separate SOC 2 Type II certifications.

### Source excerpt

Runpod has officially achieved SOC 2 Type II certification, validating that its enterprise-grade security controls not only meet strict design standards.

## Build a Halloween Murder Mystery Game with AI on Runpod

DevFeed: [Build a Halloween Murder Mystery Game with AI on Runpod](<https://devfeed.tech/articles/build-a-halloween-murder-mystery-game-with-ai-on-runpod-109145.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/halloween-mystery-game-runpod>)

Author: Jessica Garson Beauchemin

Published: 2026-10-30T17:51:56Z

Content type: tutorial

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [browser-game](<https://devfeed.tech/topics/browser-game.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [flux](<https://devfeed.tech/topics/flux.md>), [Learning](<https://devfeed.tech/topics/learning.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [build](<https://devfeed.tech/tags/build.md>), [flux](<https://devfeed.tech/tags/flux.md>), [game](<https://devfeed.tech/tags/game.md>), [halloween](<https://devfeed.tech/tags/halloween.md>), [mystery](<https://devfeed.tech/tags/mystery.md>), [python](<https://devfeed.tech/tags/python.md>), [qwen3](<https://devfeed.tech/tags/qwen3.md>)

### AI overview

The excerpt shows how to build Murder at Hollow Manor, a browser mystery where players question five AI suspects, consult a Coroner, and make one final accusation. A small Python server keeps the case and Runpod API key, enforces game rules, rate limits, and spending caps, while Runpod Public Endpoints handles inference: Qwen3 32B AWQ generates dialogue and the reveal, and FLUX.1 Schnell creates portraits that are cached. The accusation verdict is returned immediately; the reveal runs asynchronously and the browser can wait for it. The excerpt provides local setup commands and points to a deployment guide for public hosting and optional Coroner setup. It also suggests changing the setting, characters, and clues, and notes a Dockerfile for deployment.

### Source excerpt

Build Murder at Hollow Manor, a Halloween mystery game with AI suspects and generated portraits, using Python, Qwen3 32B and FLUX.1 Schnell on Runpod.

## Serverless GPU inference vs. GPU Pods: cost benchmarks

DevFeed: [Serverless GPU inference vs. GPU Pods: cost benchmarks](<https://devfeed.tech/articles/serverless-gpu-inference-vs-gpu-pods-cost-benchmarks-109146.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/serverless-gpu-inference>)

Author: Matt Sarrel

Published: 2026-10-29T23:09:35Z

Content type: comparison

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [Serverless](<https://devfeed.tech/topics/serverless.md>), [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cost](<https://devfeed.tech/tags/cost.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-inference](<https://devfeed.tech/tags/gpu-inference.md>), [inference](<https://devfeed.tech/tags/inference.md>), [pods](<https://devfeed.tech/tags/pods.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [vs](<https://devfeed.tech/tags/vs.md>)

### AI overview

On the tested A40, a busy Runpod Serverless worker cost about 2.3 times as much per output token as a Pod; flex workers were cheaper only when billed for under roughly 29 minutes per hour at October 29, 2026 prices. Queue-based Serverless endpoints added about half a second to first-token latency when warm and several seconds under load, while load-balancing endpoint latency was not measured. FlashBoot usually woke a sleeping worker in 1-2 seconds in a 15-trial rerun, but two trials required about four minutes; a new endpoint takes longer while downloading its model. Pods provide interactive shell access and persistent volume options; Serverless manages scaling but workers are stateless, and network volumes restrict placement to one data center. The source recommends benchmarking the intended endpoint against latency and capacity needs, checking billed worker time and storage constraints, and using queue endpoints for long jobs or load balancing for standard HTTP requests within its limits.

### Source excerpt

We benchmarked one A40 as a Pod and as a Serverless worker. A busy worker cost about 2.3x per token, but Serverless wins below 29 billed minutes an hour.

## Batch Inference Is Now Generally Available on Runpod

DevFeed: [Batch Inference Is Now Generally Available on Runpod](<https://devfeed.tech/articles/batch-inference-is-now-generally-available-on-runpod-105391.md>)

Original publisher: [Read original article](<https://www.runpod.io/blog/batch-inference-is-now-generally-available-on-runpod>)

Author: The Team at Runpod

Published: 2026-10-28T16:04:37Z

Content type: release

Language: en

Sources: [Runpod Blog.](<https://devfeed.tech/sources/runpod-blog.md>)

Topics: [model-serving](<https://devfeed.tech/topics/model-serving.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [GPU](<https://devfeed.tech/topics/gpu.md>), [Web APIs](<https://devfeed.tech/topics/web-apis.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [account](<https://devfeed.tech/tags/account.md>), [all](<https://devfeed.tech/tags/all.md>), [api](<https://devfeed.tech/tags/api.md>), [api-key](<https://devfeed.tech/tags/api-key.md>), [application](<https://devfeed.tech/tags/application.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [away](<https://devfeed.tech/tags/away.md>), [b300](<https://devfeed.tech/tags/b300.md>), [backlog](<https://devfeed.tech/tags/backlog.md>), [based](<https://devfeed.tech/tags/based.md>), [batch](<https://devfeed.tech/tags/batch.md>), [existing](<https://devfeed.tech/tags/existing.md>), [inference](<https://devfeed.tech/tags/inference.md>), [queue](<https://devfeed.tech/tags/queue.md>), [send](<https://devfeed.tech/tags/send.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [up](<https://devfeed.tech/tags/up.md>), [workers](<https://devfeed.tech/tags/workers.md>)

### AI overview

Runpod Batch Inference lets you submit up to 3 million requests a day to an existing queue-based Serverless endpoint, using dedicated workers so batch jobs do not take capacity from live `/run` traffic. It suits work that can wait: start times are not guaranteed, requests run in no set order, and multi-hour latency is possible. Create a batch through the console or API, then finalize it to start processing; draft batches do not run. Check request counts and statuses, inspect per-request errors, and page through results. A batch can hold up to 5,000 requests, and an endpoint can run up to 10 active batches, subject to account limits. Billing is based on compute time per request, with discounts of up to 30% depending on GPU type (20% on H100, H200, B200, and B300). Since the service provides no pre-finalization cost estimate, the excerpt recommends trying 100 requests and extrapolating.

### Source excerpt

Batch Inference is now generally available on Runpod Serverless. Send up to 3 million requests a day to an existing queue-based endpoint, on dedicated workers.

[Next page](<https://devfeed.tech/sources/runpod-blog.md?cursor=WyIyMDI2LTEwLTI4VDE2OjA0OjM3KzAwOjAwIiwgIjgwZTIwZTUxLWY4MjMtNGIzYi05MDdiLTI3ZWUzN2E3NmE3ZCJd>)