GitHub Models
Every GitHub account gets free, rate-limited API access to a large catalog of frontier and open models (GPT-5/4.1, Llama, DeepSeek, Grok, Mistral, Cohere, Phi) for prototyping.
- Requests / day
- 8
- Requests / min
- 1
- Card required
- No
GitHub Models scores 3.8/10 for free value in LLM. The main constraint: Free rate-limited access on every GitHub account; per-model request/token caps that vary by Copilot plan (Free tier has the lowest). No card or payment needed to stay on the free quota; usage blocked (not billed) once quota is hit unless you opt into paid usage.
Free value score
3.8 / 10Show the math
Key metrics
Models
30 free · 60 totalLimits may be stale — last re-verified outside this section's freshness window.
| Model | Family | AA | Access |
|---|---|---|---|
| DeepSeek-R1 | DeepSeek | 20.1 | free |
| gpt-4.1 | GPT | 19.4 | free |
| gpt-4.1-mini | GPT | 16.3 | free |
| Llama-4-Maverick-17B-128E-Instruct-FP8 | Llama | 14.3 | free |
| gpt-4o | GPT | 11.2 | free |
| DeepSeek-V3 | DeepSeek | 10.4 | free |
| Llama-4-Scout-17B-16E-Instruct | Llama | 10 | free |
| Meta-Llama-3.1-405B-Instruct | Llama | 8.5 | free |
| gpt-4.1-nano | GPT | 7.3 | free |
| gpt-4o-mini | GPT | 6.9 | free |
| Codestral-2501 | Mistral | — | free |
| DeepSeek-R1-0528 | DeepSeek | — | free |
| DeepSeek-V3-0324 | DeepSeek | — | free |
| grok-3 | Grok | — | free |
| grok-3-mini | Grok | — | free |
| Llama-3.2-11B-Vision-Instruct | Llama | — | free |
| Llama-3.2-90B-Vision-Instruct | Llama | — | free |
| Llama-3.3-70B-Instruct | Llama | — | free |
| MAI-DS-R1 | DeepSeek | — | free |
| Meta-Llama-3-70B-Instruct | Llama | — | free |
| Meta-Llama-3-8B-Instruct | Llama | — | free |
| Meta-Llama-3.1-70B-Instruct | Llama | — | free |
| Meta-Llama-3.1-8B-Instruct | Llama | — | free |
| Ministral-3B | Mistral | — | free |
| Mistral-large-2407 | Mistral | — | free |
| Mistral-Large-2411 | Mistral | — | free |
| mistral-medium-2505 | Mistral | — | free |
| Mistral-Nemo | Mistral | — | free |
| Mistral-small | Mistral | — | free |
| mistral-small-2503 | Mistral | — | free |
| o3 | GPT | 30.4 | paid |
| o1 | GPT | 23.4 | paid |
| o3-mini | GPT | 19 | paid |
| o1-preview | GPT | 17 | paid |
| gpt-5 | GPT | — | paid |
| gpt-5-chat | GPT | — | paid |
| gpt-5-mini | GPT | — | paid |
| gpt-5-nano | GPT | — | paid |
| o1-mini | GPT | — | paid |
| o4-mini | GPT | — | paid |
What you get
Free for any GitHub account with no card and no payment opt-in required
Single key reaches 40+ models incl. GPT-5/4.1, Llama 4, DeepSeek, Grok 3, Cohere
OpenAI-compatible / Azure AI Inference endpoint, easy drop-in
Usage is blocked rather than billed when the free quota is exhausted
Gotchas
No gotchas documented yet. Contribute one if you know a constraint we missed.
Sources
-
"DeepSeek-R1, DeepSeek-R1-0528, and MAI-DS-R1 have Copilot Free limits of 1 RPM and 8 RPD. The same GitHub table lists separate token-per-request caps."
https://docs.github.com/en/github-models/use-github-models/prototyping-with-ai-models -
"Azure OpenAI o1, o3, and gpt-5 | Requests per minute | Not applicable | 1 | 2 | 2 | Requests per day | Not applicable | 8 | 10 | 12"
https://docs.github.com/en/github-models/use-github-models/prototyping-with-ai-models -
"If you want to develop a generative AI application, you can use GitHub Models to find and experiment with AI models for free. Once you are ready to bring your application to production, opt in to paid usage for your enterprise."
https://docs.github.com/en/github-models/use-github-models/prototyping-with-ai-models -
"GPT-4.1 scored 66.3% on the difficult Diamond tier of questions"
https://www.rdworldonline.com/openai-claims-gpt-4-1-sets-new-90-standard-in-mmlu-reasoning-benchmark/ -
"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) currently leads the Artificial Analysis Intelligence Index with a score of 60 ... The top AI models by Intelligence Index are: 1. Claude Fable 5 (60), 2. Claude Opus 4.8 (56), 3. GPT-5.5 (xhigh) (55)"
https://artificialanalysis.ai/models/comparisons/gpt-4-1-vs-gpt-4o-2024-08-06
Alternatives
Same category · similar quota
Unlimited code completions + premium requests (legacy 300/mo, now usage-based AI Credits, 1 credit = $0.01) across GPT-5.5/5.4/Codex, Claude Opus 4.8/Sonnet 5, Gemini 3.1/3.5, and Kimi K2.7 Code (GA); agent mode in VS Code/JetBrains/CLI
Kimi K2.7 Code + K2.7 HighSpeed (~180 tok/s) + K2.6; ~300-1,200 requests/5h with weekly quota, up to 30 concurrent; OpenAI- AND Anthropic-compatible endpoints (Claude Code, OpenCode, OpenClaw, custom agents)
