Skip to content
Latent Space· swyx .ai·· 5 h ago

[AINews] Gemini 4 Carbon tops leaderboards, launches for cyber defenders

[AINews] Gemini 4 Carbon tops leaderboards, launches for cyber defenders

AI summary

Google’s new frontier model leads independent evals, with access initially limited to trusted defenders.

Selection record

AdmittedSum of both 164 ≥ twice the threshold 152

Source tier
Media and individuals; this tier's threshold is 76
Pre-filter
passed:Gemini 4 Carbon模型发布及基准评测
Why it was chosen
Rounds up where Gemini 4 Carbon places on several independent leaderboards and its pricing, and sets out the specific differences in coding, hallucination and access restrictions.
Same event

A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.

Source: Latent Space · latent.space