resolves to my impression a week or two after launch and I've had the opportunity to try it out. i should want to use it just as much as openai’s model and anthropic’s for coding mnx.fi. Flash/instant/mini doesn't count. Should be pro or similar. 3.2 pro-preview, 3.5 pro, 4 pro all okay
Update 2026-05-20 (PST) (AI summary of creator comment): Speed considerations: If a Gemini model lacks a fast tier (e.g., like Opus's fast tier) and that alone causes the creator to prefer other models, that is not sufficient to resolve NO. However, the model must still be reasonably fast to be considered frontier.
1,000I should say that it could be the case that i don't want to use the gemini model as much bc maybe it's slower than the other frontier models. I'm tempted to say that shouldn't count if the actual output (though perhaps too slow) is as good/better than the other frontier models. Any dissenters @traders?
@ian personally I would say that if the model is so slow you're unwilling to use it that should at least be some points off from it being considered 'frontier', but I can see the argument either way. We know LLM performance scales with reasoning length without really hitting a wall - if Gemini 3.5 beats other models slightly but only by reasoning for 500k tokens for every request at insane time/monetary cost, would that model really be 'frontier' in any meaningful sense? Personally I would say no. By extension if the model is generally strong but very slow I would at least consider that factor when resolving.
@2b3o4o yeah fair. it has to be reasonably fast to be considered. if opus has the fast tier and i use that and not gemini bc it doesn't have a fast tier, that shouldn't be enough to resolve no