GitHub Models gives you free, rate-limited access to 40+ AI models, with limits like about 50 requests/day for High tier models (for example, GPT-4o). If you’re searching for GitHub Models free credits, this is basically “free inference” without the usual billing setup.
Solo devs prototyping an app, startup teams trying to stretch runway, and researchers who want to compare models quickly all get something useful here. No credit card. No Azure account. Just your GitHub login.
This guide covers eligibility, the exact signup steps for playground and API access, the rate limits that matter, and a few ways to squeeze more value out of the free tier.
Program at a Glance
| Provider | GitHub |
| Credit Amount | Rate-limited free access (ex: 50 requests/day High tier) |
| Duration | Ongoing (limits reset by time window) |
| Eligibility | Any GitHub account gets free access (per-account limits). |
| Credit Card Required? | No, not for the free tier. |
| Difficulty | Easy. Sign in and use the playground immediately. |
| Best For | Prototyping, model comparisons, CI experiments in Actions |
| Official Page | GitHub Program Page |
What You Actually Get
GitHub Models gives every GitHub user free, rate-limited access to a catalog of 40+ models from OpenAI, Meta (Llama), DeepSeek, Mistral, Microsoft (Phi), Anthropic (Claude), Cohere, xAI (Grok), and AI21 Labs. You can use the in-browser Playground for interactive testing, or call the OpenAI-compatible chat completions endpoint (https://models.github.ai/inference/chat/completions) using a fine-grained personal access token with the models:read scope. GitHub also supports prompt files (.prompt.yml) for version-controlled prompts, plus evaluations for side-by-side model comparison.
In practical terms, this is enough to prototype a feature end-to-end: build a small chat flow, test a few system prompts, swap models by changing the model parameter, and wire it into a dev app or a GitHub Actions workflow. It’s not meant for heavy production traffic, but frankly it’s one of the smoother “try lots of models” setups out there.
Who Qualifies (and Who Doesn’t)
Eligibility is simple: if you have a GitHub account, you can use GitHub Models. The free tier is per account and comes with strict rate limits that may vary based on your GitHub Copilot plan.
- You need a GitHub account (free accounts work).
- No credit card or payment method is required to use the free tier.
- API access requires a fine-grained personal access token with
models:readscope (required since May 2025). - GitHub Actions usage works best when you add
models: readto workflow permissions and use the built-inGITHUB_TOKEN.
If you’re trying to “stack” free limits across multiple tokens or repos under one account, that won’t work. Limits apply per GitHub account, not per token or project.
How to Sign Up
Playground access takes about 2 minutes; API access takes a little longer because you need to generate a token.
- Go to github.com/marketplace/models.
- Sign in with any GitHub account (free accounts work).
- For the Playground, click any model to open an interactive chat playground (it works immediately, with no setup).
- For API access, go to Settings → Developer settings → Personal access tokens → Fine-grained tokens.
- Generate a new token with the
models:readscope. - Use the token with the API endpoint https://models.github.ai/inference/chat/completions.
- Set the
Authorization: Bearer <YOUR_TOKEN>header in your requests.
Once you’re signed in, the Playground works right away. For API calls, there is no separate “approval” step; if your token has models:read and you’re under the limits, requests go through.
What the Credits Cover
GitHub Models isn’t a cash credit program. It’s free inference with rate limits, covering an OpenAI-compatible chat completions API, a browser Playground, and workflow tooling like prompt files and evaluations. You can also list models via the REST catalog endpoint (GET /catalog/models).
| Service / Feature | What It Does | Included? |
|---|---|---|
| Playground | In-browser chat UI to test models and prompts. | ✓ |
| Models inference API | OpenAI-compatible chat completions endpoint with PAT auth. | ✓ |
| Prompt files + evaluations | Version-controlled prompts and side-by-side comparisons. | ✓ |
| Pay-as-you-go + BYOK | Scale via GitHub billing or connect your own provider key. | Partial |
Don’t assume “free” means “long context.” The free tier has an 8K input and 4K output token cap per request, so big-document summarization is a bad fit.
Limitations to Know About
Every free program has catches. With GitHub Models, the catch is strict rate limits and per-request token caps that are easy to hit if you build anything interactive.
- Rate limits are enforced per model tier (Low, High, or Embedding), not “one global pool.”
- High tier examples like GPT-4o, GPT-5, and o3 are limited to about 10 requests/min and 50 requests/day.
- Low tier models (like Phi-4, Llama 3.1 8B, and Ministral 3B) are higher at about 15 requests/min and 150 requests/day.
- Token caps are per request (about 8K input and 4K output), so you can’t “batch” around them.
- When you hit a limit, calls return HTTP 429 until the window resets.
- Limits apply per GitHub account, not per token or project, and they may change without notice.
- Your GitHub Copilot plan can affect the exact rate limits, even though all accounts get access.
- Calls from GitHub Actions count against the same account limits as direct API calls.
When you run out of free allowance, nothing “breaks” permanently. Requests just fail with HTTP 429 until the reset. If you need more throughput, you can opt into pay-as-you-go billing through GitHub, or switch to BYOK so your provider subscription governs rate limits instead of GitHub’s free tier.
Have Unused GitHub Credits?
A lot of teams end up with credits they can’t use in time, especially when they get them through bundles, promos, or company agreements. Those balances tend to expire quietly, which is painful when budgets are tight. If you’re sitting on unused GitHub credits you won’t burn down, AI Credit Mart lets you list them and sell at up to about 70% of face value instead of watching them go to zero.
Need More GitHub Credits?
If you outgrow the free tier, paying retail isn’t your only option. AI Credit Mart has discounted GitHub credits from companies with surplus allocations, often priced 30–70% below face value. It’s a clean way to extend your runway while you decide whether to go pay-as-you-go or BYOK for production.
Tips for Getting the Most Out of Your Credits
- Start in the Playground first, because it works instantly and helps you avoid burning requests on basic prompt debugging.
- Keep an eye on model tiers (Low vs High vs Embedding) on each model’s marketplace page so you don’t accidentally prototype on a 50-requests/day model.
- Reuse your existing OpenAI chat-completions code, since the GitHub inference endpoint uses an OpenAI-compatible format and often needs minimal changes.
- Switch models by changing only the
modelparameter, which makes A/B comparisons fast in the same endpoint and auth setup. - If you’re calling models from CI, add
models: readto the GitHub Actions permissions block and use the built-inGITHUB_TOKEN, but remember Actions usage counts against your account rate limits.
Frequently Asked Questions
There isn’t a fixed dollar credit here; the “value” is free, rate-limited inference, like about 50 High-tier requests/day (for example GPT-4o) or about 150 Low-tier requests/day (for example Phi-4). Practically, that’s plenty for prompt iteration, basic evaluations, and a small prototype that only a few people touch. Once you start running automated jobs or team-wide traffic through it, you will hit 429s fast. At that point, switch to pay-as-you-go or BYOK.
No.
The free tier is ongoing, but usage is limited by per-minute and per-day request windows (and a per-request token cap). When you hit the limit, you get HTTP 429 responses until the window resets.
Yes. If you have GitHub credits you won’t use before they expire, you can list them on AI Credit Mart and sell them at up to 70% of face value. Companies regularly list surplus credits from startup programs and enterprise agreements.
AI Credit Mart has discounted GitHub credits available from companies with surplus allocations. Prices are typically 30-70% below retail.
For GitHub Models’ free tier specifically, you don’t get charged; you just get HTTP 429 until limits reset, or you upgrade to pay-as-you-go/BYOK for higher throughput.
You need a fine-grained personal access token with the models:read scope (required since May 2025).
No. Calls from GitHub Actions count against your account’s rate limits, so a busy workflow can eat the same daily allowance you planned to use for manual testing.
GitHub Models’ free tier is real utility: instant playground access and an OpenAI-style API across a surprisingly large model catalog. Use it to prototype fast, then scale with pay-as-you-go, BYOK, or discounted credits if you want to avoid full-price burn.
Your AI credits are losing value every day
Join the marketplace and start trading unused credits today.