AI inference endpoint model availability question

[hphan508]

Select Topic Area

Question

Body

I am using this is action
uses: actions/ai-inference@v2.0.5
with:
model: openai/gpt-5
endpoint: https://models.github.ai/orgs/${{ github.repository_owner }}/inference

i would get an error saying "simpleInference: chatCompletion failed: Error: 400 Unavailable model: gpt-5" . I see the model gpt-5 as enable in my org. In this endpoint, gpt-4o is available. Do anyone have an idea why is this the cases?

[kmoragap]

Your org inference endpoint only supports models that show up in the Models Catalog for your org. Check GET /catalog/models and use the exact model id listed there; if openai/gpt-5 isn’t in the catalog yet, use something that is (e.g. openai/gpt-4o) until it’s rolled out.

Docs link (catalog endpoint):
ttps://docs.github.com/en/rest/models/catalog
​

[hphan508]

If i am the owner of my org, how do i make the model available? Within the webgui portal i do see the gpt5 as enable and available.

[kmoragap]

seeing GPT-5 "enabled" in your org web UI doesn't mean it's live in the inference endpoint yet. The API only recognizes models actually in your Models Catalog.

Check what's actually available at https://models.github.ai/catalog/models.

If openai/gpt-5 isn't in that catalog response, it's not rolled out to your org yet. Use openai/gpt-4o or whatever models DO appear until GPT-5 shows up.

[pranitacharjunewar3-sketch]

This error happens because the inference endpoint you are calling does not support the gpt-5 model, even if it’s enabled in your org. GitHub’s models availability can differ per endpoint and per action version.

Try using a model that is supported by that endpoint (for example gpt-4o) and confirm available models for your endpoint. You can list models from the inference endpoint here:

https://docs.github.com/en/rest/actions/ai-inference#list-models

Also make sure your workflow uses the latest version of actions/ai-inference which supports newer models.

[monxcode]

Error Explanation:
You are getting
"simpleInference: chatCompletion failed: Error: 400 Unavailable model: gpt-5"
This happens because the GitHub inference endpoint you are using
does not currently support the model gpt-5 even if it is enabled in your organization

Reasons:

1. The endpoint only exposes certain models such as gpt-4o at this time
2. Some GPT-5 variants may require a newer endpoint or updated API version
3. The rollout of GPT-5 support may not be complete for your organization endpoint

How to fix or troubleshoot:

Step 1: List available models from your endpoint
curl -H "Authorization: token GITHUB_TOKEN"
https://models.github.ai/orgs/your_org/inference/models

Step 2: Use a supported model from the list, for example
model: openai/gpt-4o

Step 3: Update your GitHub Actions workflow
uses: actions/ai-inference@v2.0.5
with:
model: openai/gpt-4o
endpoint: https://models.github.ai/orgs/github.repository_owner/inference

Step 4: Check GitHub documentation and changelog for when GPT-5 becomes fully available
https://docs.github.com/en/rest/models/inference

Summary:
The error is because gpt-5 is not yet available on that endpoint
Use gpt-4o or any model returned by the /models API call
Keep an eye on GitHub documentation for GPT-5 rollout updates

[its-Kashie]

Select Topic Area

Question

Body

I am using this is action uses: actions/ai-inference@v2.0.5 with: model: openai/gpt-5 endpoint: https://models.github.ai/orgs/${{ github.repository_owner }}/inference

i would get an error saying "simpleInference: chatCompletion failed: Error: 400 Unavailable model: gpt-5" . I see the model gpt-5 as enable in my org. In this endpoint, gpt-4o is available. Do anyone have an idea why is this the cases?

What’s really happening

Even though gpt-5 shows as “enabled” in your org, that does NOT mean it’s available on every inference endpoint.

You’re calling this endpoint:

https://models.github.ai/orgs/${{ github.repository_owner }}/inference

That endpoint is GitHub Models–scoped, not OpenAI-global.
Right now, GitHub Models only exposes a limited, curated set of models, and gpt-5 is not one of them yet.

So the error:

400 Unavailable model: gpt-5

means:

❌ The endpoint exists
❌ Your org is valid
❌ The model is enabled in theory
✅ But this endpoint does not serve gpt-5

---

Why gpt-4o works but gpt-5 doesn’t

GitHub Models currently supports things like:

• gpt-4o
• gpt-4o-mini
• some embeddings / text models

But gpt-5 is still restricted to OpenAI-native endpoints and is not yet wired into GitHub Models inference.

Org enablement ≠ endpoint availability.

---

How to confirm this yourself

You can list models available to GitHub Models like this:

curl https://models.github.ai/orgs/<ORG_NAME>/models \
  -H "Authorization: Bearer $GITHUB_TOKEN"

You’ll notice gpt-5 is missing.

---

What you can do right now

✅ Option 1: Use a supported model (recommended)

uses: actions/ai-inference@v2.0.5
with:
  model: gpt-4o
  endpoint: https://models.github.ai/orgs/${{ github.repository_owner }}/inference

This is the only combo that works today.

---

✅ Option 2: Use GPT-5 via OpenAI directly (not GitHub Models)

If you must use GPT-5, you need:

• OpenAI’s API endpoint
• An OpenAI API key
• Not GitHub Models

Example (outside GitHub Models):

https://api.openai.com/v1/chat/completions

---

TL;DR

• ✅ gpt-5 enabled in org → true
• ❌ gpt-5 available on GitHub Models inference → false
• ✅ gpt-4o works → expected
• ❌ This is not a config bug or permission issue

GitHub just hasn’t exposed GPT-5 on that endpoint yet.

[kyim120]

Thanks I Got MIne Issue

[PythonPlumber]

If you see the model enabled in your org settings but the action keeps spitting out a 400 error, it usually feels like you're being ghosted by the API.

The main thing to keep in mind is that the model name in the GitHub UI doesn't always match the exact string the inference endpoint expects. Even if your org has access to a specific model, the endpoint might be looking for a very specific version name.

Here is what is likely going on and how to poke it to see what's actually available:

First, try changing the model string. Usually, GitHub's inference service prefers specific versions over the generic name. Instead of openai/gpt-5, try using gpt-5 or whatever the specific version string is listed in the GitHub Models marketplace. Since gpt-5 is still very new or in limited release depending on your tier, it might actually be mapped under a different identifier like gpt-5-preview or something similar.

Second, you can actually ask the endpoint what it has in stock. If you have a PAT (Personal Access Token) with the right permissions, you can run a quick curl command against the models list endpoint:

curl -H "Authorization: Bearer YOUR_TOKEN" https://models.github.ai/inference/models

This will give you a JSON list of every single model ID that your specific token and org can actually call. If gpt-5 isn't in that list, then the "enabled" status in the org settings hasn't synced to the inference runtime yet.

Third, double check your permissions in the action. Ensure the token being used by actions/ai-inference actually has the models:read or equivalent scope for the org. Sometimes the default GITHUB_TOKEN in actions doesn't have the "juice" to see the high end models even if the org owner enabled them.

If gpt-4o works but gpt-5 doesn't with the exact same setup, it’s almost definitely down to the specific model ID string or a rollout delay on GitHub's backend.

Give that curl command a shot to see the "true" list of available IDs! Let me know what it returns and we can narrow it down from there.

[pratikrath126]

For model availability at the AI inference endpoint, the most reliable source is the GitHub Models API documentation and the Model Catalog. You can programmatically list the available models using the following GET request:

(Note: You'll need a valid GitHub token with the appropriate permissions).
Additionally, keep an eye on the GitHub Marketplace and the Models section in your GitHub settings, as new models are added periodically during the preview phase.