N/A if Gemini 3.0 Pro Preview is never released
"Gemini 3.0 Pro experimental" also counts
Will resolve according to a poll of 2-3 uninvolved mods who will be given a case for each position, if ambiguous
Update 2025-11-06 (PST) (AI summary of creator comment): Rate limits will be considered very low if they are lower than Gemini 2.0 Flash Thinking's rate limits at the time of Gemini 3.0 Pro Preview's release.
Latency will be considered high if it is higher than Gemini 2.0 Flash Thinking's latency at the time of Gemini 3.0 Pro Preview's release.
Multiple unstable weights means there are at least 2 different model weights/versions available simultaneously (e.g., different experimental versions or snapshots).
1,000People are also trading
Feel like this should be no, but comparing it against 2.0 flash thinking? like latency according to openrouter? Time between question and response? 3.0 pro's would obviously be higher than the flash model. Since they're both thinking models and 3.0 would be much bigger. And also I don't think 2.0 thinking is even being served anymore, I don't see it on openrouter or aistudio. 2.5 flash thinking is though.
@Bayesian idk what would work, but I'm not gunna bet heavy cause polls seem so unpredictable on this
@AffineTyped i'll ask my elected representative to raise taxes on bullshit
@Bayesian fwiw I think the format that works the best is something like "If Google or Logan explicitly mention any of these things OR if the final release model has more preferable parameters on this front than preview