Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaks

Gemini 4 Carbon debuts at #1 on Artificial Analysis' new index, excels at finding and patching vulnerabilities, and expands Fairwind Program access ahead of wider Antigravity rollout.

A post says Gemini 4 Carbon is finally here and beats Opus 5.5 on almost every benchmark tested, asking whether the results are real or just benchmark-maxxed.

A user says Gemini 4 Carbon is cheaper than Opus 5.5 and beats it on almost all benchmarks, putting pressure on Anthropic, and asks for access.

Gemini 4 Carbon reportedly scores 59 on the AA Intelligence Index, one point above Opus 5.5 and two above Astra, which the poster calls a nightmare for Anthropic.

AI_Screening posts a video said to be Gemini 4 Carbon's output when it appeared in Arena as gemini 4 argon, prompted with a Bentley convertible at sunset.

The poster argues Gemini 4 Carbon is what Google originally intended to release as Argon, and that its shipping problem now looks solved.

Gemini 4 Carbon scores 54.8% on ExploitGym versus 42.4% for GPT-6 Astra and 34.5% for Opus 5.5, and also leads on CWE-Bench and Sec-Bench.

Two weeks after Argon, Google releases Gemini 4 Carbon, ranking first on the new Artificial Analysis index ahead of Opus 5.5 with strong cyber vulnerability results, first to trusted defenders.

A claim that Gemini 4 Carbon shipped two weeks after Argon, already ranks #1 on Artificial Analysis, beats rivals on coding and long tasks, and reaches security teams first.

A leaked table claims Gemini 4 Carbon leads 15 of 17 benchmarks against GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon, scoring 59 on the AA Intelligence Index with Argon-like pricing.

Gemini 4 Carbon leads on security benchmarks (88.2% SEC-bench Pro, 71.0% CWE-bench), rolls out first to cyber defenders at $2/$10 per 1M tokens, and scores 96.2% on GraphWalks long context.

Two weeks after Gemini 4 Argon, a Google staffer says the next model, Gemini 4 Carbon, has good internal vibes and will roll out to cyber defenders before wider access.