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

Google says Carbon scores 59 on Artificial Analysis' Intelligence Index, one above Opus 5.5, wins 14 of 16 benchmarks, and is limited to select cyber defenders for now.

Gemini 4 Carbon reportedly stealth-tested on Arena for days and now leads Artificial Analysis at 59 over Opus 5.5's 58, though early testers flagged slow speed.

A Finnish-language post claims Google's new Gemini 4 Carbon model beats Claude Opus 5.5 and GPT-6 Astra on benchmarks.

fofrAI shares the Google blog link announcing Gemini 4 Carbon, quoting the company on pushing the frontier faster while strengthening safeguards.

A user posts Gemini 4 Carbon benchmark screenshots, saying it crushes Claude Opus 5.5 and GPT-6 Astra and that Google is back.

The post claims Google launched Gemini 4 Argon and that it scores higher than GPT-6 Astra and Opus 5.5, with a screenshot attached.

A video post lists Gemini 4 Carbon scores including 59 on AA Intelligence Index, 88.2% on SEC-bench Pro and 65.2% on Terminal-Bench 4.0, claiming wins over GPT-6 Astra.

Gemini 4 Carbon scores 59 on the AA Intelligence Index, topping Claude Opus 5.5 and GPT-6 Astra, and leads on finding, exploiting and patching flaws, with access restricted to select security groups.

A user guesses Gemini 4 Carbon is a new Argon checkpoint and questions why it isn't called Gemini 4.1 or how Carbon relates to Argon.

A poster claims Gemini 4 Carbon outperforms Opus 5.5, Fable 5.1 and GPT-6 Astra, just 15 days after Gemini 4 Argon.

A short opinion post ordering Gemini 4 Carbon ahead of Opus 5.5 and GPT-6 Astra, with no benchmarks or evidence given.

Andon Labs reports Gemini 4 Carbon ranks first on Blueprint-Bench 2, with AI now tying humans on physical 3D understanding benchmarks.