Gemini 3.1 Pro – just launched by Google
Google just launched its latest Gemini 3.1 Pro model and it's looking good on the evals so far. When you see the below chart, you will notice that it's being shown far better than even Opus 4.6 and GPT-5.3-Codex models.
As mentioned in their launch announcement, Gemini 3.1 Pro is great at creating coding, code-based animations, complex system synthesis, interactive designs, etc. The new model is already available via the API and in Gemini Web, Antigravity, AI Studio, Gemini CLI, etc.
Currently, I am not seeing a new Gemini 3.1 Pro option in the Gemini app, but I think the default Pro option in the app is new model.
I am still testing it, and will keep updating this post as I discover more.
Update:
I can see the Gemini 3.1 Pro model available in AI Studio and I can use the model as well.
Update:
I have been using the new model for the last few years and I think it's only better in terms of design, but still not good at coding. The GPT-5.3-Codex and Opus 4.6 are far better models in terms of coding.
Update:
I tried using the new model again, and it's not good, honestly. Only does good design and animations, but feels really dumb to use and often times introduces lots of bugs in the code, forgets things, and is bad at tool calling as well.
I don't know how it scores so much on evals.
Webmentions