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

The author compares Kimi K3.1 Pro (max) against Gemini 4 Carbon (high), noting K3.1 Pro's output reminds him of Astra.

On the same prompt Kimi cost $0.75 versus $2.30 for Gemini 4 Argon, with the author calling quality comparable at a fraction of the price.

A post relays Sundar saying Gemini 3.5 Pro was dropped because Argon pretraining had finished by the time the Flash-first playbook was ready for Pro.

On Lech Mazur's extended NYT Connections benchmark, Kimi K3.1 Pro edges below K3 with fewer reasoning tokens, while Gemini 4 Argon (high) still ranks first over GPT-6 Astra and GPT-6.1 Sol.

A complaint post lists Opus 5.5, Gemini 4 Argon and Kimi K3.1 Pro arriving back to back, saying OpenAI needs to wake up instead of shipping resets and bug fixes.

A joke post claims Gemini 4 Argon described GPT-6 as a bland, limp, waterlogged zucchini — safe but tasteless.

The author says Gemini 4 Argon wrote the paper's first draft and figures, then rewrote it in his style after being fed his past first-author papers.

A reply ranks computer-use price-performance as Kimi K3.1 > Gemini 4 Argon > Muse Wonder > Sol 6.1.

On the Physical-Engineering's Last Exam leaderboard, Claude Opus 5.5 leads at 48.3%, GPT 6 Astra at 45.8%, Gemini 4 Argon at 40.6%, and Grok 4.7 at 30.1%.

A Kimi Desktop test generated a pelican riding a bicycle, which the author says is about as good as Gemini 4 Argon, minus the nitro boost.

Moonshot's Kimi K3.1 Pro scores 55 on the Artificial Analysis Intelligence Index, behind Claude Opus 5.5 and Sonnet 5.5, while Gemini 4 Argon sits at 53.

A developer connected their OKF Brain to an algo-update MCP and Gemini 4 Argon, saying the combo outperforms them alone, alongside health and voice-clone apps.