Post by @lord.luke5

Gemini 4 Argon is Google's new top AI model, and on paper it beats GPT-6 and Claude Gemini 4 Argon is Google's new frontier AI model. On Google's own benchmark chart it takes first or ties for first on 14 of 19 tests against GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1 What it's actually good at, in plain words: Business work: running long multi-step tasks, finance research and legal writing. On the legal test it scored 19.6% while everyone else was under 7% Coding: tops DeepSWE and Vibe Code Bench, so building apps and websites from a prompt Huge amounts of info: it can keep track of up to about a million tokens, basically a whole book or codebase, and it can write answers way longer than before Video and charts: best at understanding long videos and reading charts Security: trained to find and fix bugs in software Where it doesn't win: GPT-6 still leads some coding and computer-use tests, and Claude Opus 5.5 leads Terminal-bench and ML engineering The catch: you can't use it yet. It's going to cybersecurity partners first, then paid subscribers and the API [your take here, 1 line] Is this Google's comeback or just a good-looking chart? Gemini 4 Argon, Gemini 4, Google Gemini, Google AI model, best AI model, AI benchmarks, Gemini vs GPT-6, Gemini vs Claude, GPT-6 Astra, Claude Opus 5.5, AI for coding, vibe coding, long context AI, new AI model, AI news