Create a list of officially supported models / tools

Overview

Some bugs such as #3557337: Request to create page from uploaded image , triggeres all the sub agents are caused by smaller models, its not reasonable to expect something as complex as Canvas AI to run on an OSS model on your phone.

I've set to major as it helps reduce the scope of which bugs we need to fix.

Proposed resolution

We maintain a list of models we officially support (and test against), list of models we may support, and example models we don't support. Note: The usage tier is very important. We could put it in a readme file?

I've given reasons for the examples of not supported. My proposal is that we put this in the readme file to help people make descisions when they try models out not on this list. However we could move that rationale to maybe a documentation page?

Supported

• Claude Opus 4.x - Tier 2+. (There is one tier below this)
• Chatgpt 5.1+ - Tier 2+ (There are two tiers below this)

Likely Supported

• Gemini Pro 3
• Mistral Large

Example of not supported

• Chatgpt on the free tier - Rate is too low, we might be able to do variable rates but the Agents will be incredibly slow
• Grok 3 - This will be good enough but we need to work on our Grok provider or move to Symfony AI
• Sonnet 3.7 - This is getting deprecated)
• Sonnet 4.x / Chatgpt 5 mini - This may work but we won't be focusing on targetting it.
• Gemini 3 Flash - Not supported but a good chance it will compared to the other smaller versions of frontier models and so maybe if we test this we can move this up.
• Llama 3 - Not yet found an OSS model that works well but we are looking.

User interface changes