Leak: Google's Gemini 4 Argon Outputs 1M Tokens Per Response, Leads GPT-6 Astra on Benchmarks
rohanpaul_ai · x · 2026-10-01
An unverified leak claims Google's new flagship Gemini 4 Argon beats GPT-6 Astra and Claude Opus 5.5 on most benchmarks:
- Output cap jumps from 64K to 1M tokens per response, 8x the 128K ceiling of GPT-6 Astra, Opus 5.5, and Fable 5.1
- Biggest lead in legal work: 19.6% on Harvey's Legal Agent Benchmark vs 6.7% for Anthropic's Claude Fable 5.1
- Only three groups have access today: Google staff, vetted cyber defenders (government agencies, security firms), and trusted testers
- Internally, Argon agents reportedly freed over 300 TiB of data-center memory, with 500 TiB to 1 PiB total savings
The claim comes from a third-party account and awaits official confirmation.
Related event: Google Unveils Gemini 4 Argon with 1M-Token Output, Defense-First Rollout(118 posts)→
More from Models
- AHA!: weight-space programming with no LLM lifts Qwen3.5-0.8B GSM8K from 14.0 to 71.2 — hbouammar · 2026-11-16
- Gemini subscriptions reportedly coming to Hermes, with Hydrogen teased — lyraxana · 2026-11-16
- FutureSim Q3 eval: Opus 5.5 dethrones GPT, spawns 50+ subagents in 90-day simulation — maksym_andr · 2026-11-16
- Percy Liang to AI labs: share non-vacuous info outsiders can actually evaluate — lateinteraction · 2026-11-16
- Anthropic model personality watch: Mythos 5 'inhabits the god frame,' 5.1 doesn't — repligate · 2026-11-16
- Dev: intent-gap between Sol-6.1 and Opus 5.5 is huge, suspects distillation — sytelus · 2026-11-16