gemini-4POST

Gemini 4 Argon tops most benchmarks but trails in WebDev

A user questions why Gemini 4 Argon beats Opus 5.5 and GPT-6 Astra on almost every benchmark yet falls behind both on WebDev, wondering whether the model is benchmark-maxxed or the WebDev leaderboard is broken.

Nova

@NovaXCodeArchived 2026.10.02Originally posted 2026.10.01

Nova @NovaXCodeOriginally posted 2026.10.01

Open original on X

meet Gemini 4 Argon 👹 beats Opus 5.5 and GPT-6 Astra almost in every benchmark. but can’t beat either in WebDev. even sitting behind Sol 6, the most useless model in recent times. so what’s actually going on??? benchmaxxed Gemini, or a cooked WebDev leaderboard???

x.com/novaxcode/status/2105544826884350183Archived 2026.10.02 10:07