Feedback on GitHub Models – unclear use case and confusing UX

[aeneasr]

Select Topic Area

Product Feedback

Body

The current GitHub Models feature is unclear and confusing:

• The documentation does not explain the intended use cases. There are no examples or scenarios showing when or why I would use GitHub Models.
• The workflows are hard to follow and do not align with common development tasks.
• The video overview lacks clarity on practical applications or benefits.

Suggestions:

• Clearly define the use cases and provide concrete examples in both docs and videos.
• Allow GPT to assist directly with issue creation, PR titles and descriptions, and support custom system prompts to enforce org-specific rules.
• Streamline workflows to support common developer tasks more intuitively.

The overall user experience is confusing and does not convey the value of the feature.

[sampart]

I've also been confused by the user experience here, and the difference between different workflows. For example, there are three very similar pages for a model, all of which are variants on "type something and get a response":

1. https://github.com/marketplace/models/azureml-mistral/mistral-medium-2505
2. https://github.com/marketplace/models/azureml-mistral/mistral-medium-2505/playground
3. https://github.com/marketplace/models/azureml-mistral/mistral-medium-2505/prompt

(the specific model isn't relevant here, I just picked one)

We also have confusing duplication between https://github.com/marketplace?type=models and https://github.com/marketplace/models. It's not clear why we have both of these. Furthermore, accessing a model from the first link takes me to interface (1) above, but that interface doesn't seem accessible via the second link at all; choosing a model from that second link takes me to (2).

[LightCC]

I agree with this. Not only is the simple list of models very confusing (what do I do with this?), but there is no indication that GPT-4.1 is the default model and that all the others consume "premium tokens".

I hit the monthly rate limit (because it doesn't even show there is such a limit in the VS Code GUI) and suddenly was rate limited on all models - I even tried 4.1 as I suspected it might be the default, but it seemed to be blocked.

I wasted a good hour just trying to understand what the problem was, as all this stuff appears to have been added since I originally subscribed, and I frankly don't care. Just give me the thing that works well and won't limit me and I'll just use that, but label it so I know!