GBENCH: OpenAI's GPT-6 Astra tops unsaturated coding benchmark with 83.3, sweeping top 3
teortaxesTex · x · 2026-11-11
leolinsky ran Fable 5.2 and rivals through 100 unsaturated coding/engineering environments (spatial reasoning, planning, games, circuit design). The GBENCH leaderboard shows OpenAI sweeping the top 3: GPT-6 Astra (83.3), GPT-6.1 Sol (82.0), GPT-6.1 Terra (77.2), with Claude Opus 5.5 fourth at 76.7, Gemini 4 Argon sixth (71.6), and Claude Fable 5.2 seventh (71.4).
- Author argues OpenAI's edge on open-ended technical problems explains why most mathematical breakthroughs come from OpenAI
- Anthropic is characterized as all-in on a great coding assistant rather than the smartest model
- Fable 5.2 notably pauses agentic coding runs to ask humans for clarification before tuning further
More from Models
- Astra's Hyra 2.0 trains protein models with SOTA scores and 57× speedup — zdhnarsil · 2026-11-11
- Gemini 4 Argon safety report appears to describe CoT training, contradicting GDM's claims — S_OhEigeartaigh · 2026-11-11
- Meta's Muse Spark Open, 816B params with 1M context, goes live on ai& inference — DavidBennett__ · 2026-11-11
- Fable 5.2 cost per task doubles to $26.54, undercutting Anthropic's efficiency claims — haider1 · 2026-11-11
- Carbon tops Abundant Labs agent benchmark with 14/20 vs Opus 5.5 and GPT-6 — amankhan · 2026-11-11
- Fable 5.2 called an incremental, token-hungry update; Astra could retake lead — bindureddy · 2026-11-11