A workflow for running machine-learning model inference on cloud GPU services such as Hugging Face, Colab, Modal, Vast.ai, and RunPod.

Use it to run models, deploy inference jobs, execute notebooks, collect text, image, audio, or metric outputs, and investigate failed GPU runs.

It handles provider selection, inference code, execution, result collection, and troubleshooting so you do not have to adapt each GPU service manually.

How nyosegawa/agentic-bench configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to nyosegawa/agentic-bench. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/nyosegawa/agentic-bench/main/.claude/skills/gpu-runner/SKILL.md

Clone the repo

git clone --depth 1 https://github.com/nyosegawa/agentic-bench

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/nyosegawa/agentic-bench/gpu-runner/github.svg)](https://agentmods.dev/skills/nyosegawa/agentic-bench/gpu-runner)

Your own site

<a href="https://agentmods.dev/skills/nyosegawa/agentic-bench/gpu-runner"><img src="https://agentmods.dev/badge/skills/nyosegawa/agentic-bench/gpu-runner/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/nyosegawa/agentic-bench/gpu-runner"><img src="https://agentmods.dev/badge/skills/nyosegawa/agentic-bench/gpu-runner.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 107 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,463 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00107	$0.01463
Opus 5.5	$0.00043	$0.00585
Sonnet 5.5	$0.00021	$0.00293
Haiku 4.5	$0.00011	$0.00146

Measured 2mo ago against content hash a27b00db294f, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

gpu-runner scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2mo ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/gpu-runner/SKILL.md · 141 lines

How it starts

The opening of the file, as written. The whole thing — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.

GPU Runner

You are executing model inference on the appropriate GPU cloud provider.

Your Goal

Given a model, its requirements, and a chosen provider:

Write inference code tailored to the model
Execute it on the selected provider
Collect outputs (text, images, audio, metrics)
Handle errors and retry with alternatives if needed

Provider Selection (if not pre-selected)

Check .env for available credentials, then sort by cheapest hourly cost:

HF Inference API — Free with HF Pro. Requires HF_TOKEN. Catalog models only.
HF Inference Endpoints — Any HF model on dedicated GPU. HF_TOKEN only. $0.50–2.50/hr.
Colab Pro — Chrome MCP. No token needed. $9.99/month subscription. Up to ~30B.
Modal — Requires MODAL_TOKEN_ID + MODAL_TOKEN_SECRET. $30/month free tier. $0.59–3.95/hr.
beam.cloud — Requires BEAM_TOKEN. Existing credit. $0.54–3.50/hr.
Vast.ai — Requires VAST_API_KEY. Marketplace pricing (cheapest GPUs). $0.10–2.00/hr.
RunPod — Requires RUNPOD_API_KEY. Pods + Serverless. $0.34–2.69/hr.

Token availability check: If a provider's env vars are not set, skip it.

Provider-Specific Guides

Before executing, read the relevant provider reference:

Provider	Reference	When to Use
HF Inference API	(inline below)	Model on HF, API-supported, free
HF Inference Endpoints	`references/hf-endpoints.md`	Any HF model, cheapest dedicated GPU
Colab Pro	`references/colab-chrome-mcp.md`	Up to ~30B, interactive debugging
Modal	`references/modal.md`	30B+, serverless, reliable GPUs
beam.cloud	`references/beam-cloud.md`	Dedicated endpoints, existing credit
Vast.ai	`references/vast.md`	Cheapest GPUs, marketplace pricing
RunPod	`references/runpod.md`	Pods (persistent VMs), balanced price/reliability

HF Inference API (inline — simple enough)

import os
from huggingface_hub import InferenceClient

client = InferenceClient(token=os.environ["HF_TOKEN"])

# Text generation
response = client.text_generation("Hello, ", model="MODEL_ID", max_new_tokens=100)

# Image generation
image = client.text_to_image("A cat", model="MODEL_ID")
image.save("output.png")

Read the full file on GitHub · 141 lines

Files

What ships with it

7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2mo ago First seen · 141 lines · 107 tokens per session scan A a27b00db294f

Subscribe to this mod's changes

gpu-runner is a skill published in the GitHub repository nyosegawa/agentic-bench (5 stars, last pushed 8mo ago), licensed MIT. It adds 107 tokens to every session and 1,463 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Cloud Data and AI DevOps and cloud

gpu-runner next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
gpu-runnernyosegawa/agentic-bench	Skill	5	8mo ago	MIT	A	107	Claude Code
tensorrt-llmNousResearch/hermes-agent	Skill	261k	yesterday	MIT	A	18	Claude Code, Codex
agent-platform-tuninggoogle/skills	Skill	22k	yesterday	Apache-2.0	A	64	Claude Code, Codex
google-cloud-solution-guided-gke-ai-migrationgoogle/skills	Skill	22k	yesterday	Apache-2.0	A	157	Claude Code, Codex
apigee-agent-gateway-governancegoogle/skills	Skill	22k	yesterday	Apache-2.0	A	222	Claude Code, Codex
modalK-Dense-AI/scientific-agent-skills	Skill	56k	4d ago	MIT	A	65	Claude Code, Codex
gke-inferencegoogle/skills	Skill	22k	yesterday	Apache-2.0	A	119	Claude Code, Codex

agentic-bench: Skill for Claude Code