GPT-6.2 Astra Formalizes Never-Before-Formalized 2006 Theorem: 1,794-Line Lean Proof in 94 Minutes for $21.58
Dr_Singularity · x · 2026-11-19
Vals AI reports that OpenAI's GPT-6.2 Astra autonomously formalized Smoktunowicz's 2006 Gap Theorem (no graded domain has Gelfand–Kirillov dimension strictly between 2 and 3) on its new Autoform-Hard benchmark — a theorem that, as far as they know, had never been formalized.
Key details:
- Setup: Astra worked alone in a sandbox with only Lean, Mathlib and a shell, under limits of 3 hours and a $200 budget, no human intervention. The run took 94 minutes, 1,794 lines of Lean, 146 agent turns, and cost just $21.58.
- Different route: After a shared opening, Astra's argument takes a different path from the published proof.
- Self-recovery: At 47 minutes it submitted a partial proof (complete for the commutative case, with a gap in the general one). After rejection, it filed an honest "unsolved" report, was sent back to work, thought silently for 9 minutes, then pursued a new approach — accepted 39 minutes later.
- Leaderboard: Of 56 research-level formalization tasks, GPT-6.2 Sol and Claude Fable 5.2 each solve 48; Astra solves a 49th — this theorem — and is the only model to do so.
Related event: GPT-6.2 Astra Formalizes 2006 Theorem in 94 Minutes for $21.58(6 posts)→
More from Models
- Doctor says worrying about LLM citation hallucination means you're 2 years behind — flngr · 2026-11-19
- Mystery stealth model appears on benchmark leaderboard, beating Fable 5.2 — gajus0 · 2026-11-19
- Gemini Neon reportedly matches Opus quality with 3-4x the quota of Claude or Codex — s_scardapane · 2026-11-19
- Frontier ranking debate: OpenAI credibly third for the first time as Anthropic's lead narrows — teortaxesTex · 2026-11-19
- Microsoft Claims Its Router Is '79x Cheaper' Than GPT-Sol 6.1—By Its Own Test — YvesMulkers · 2026-11-19
- Reddit user quits Gemini after Argon replaced by lesser Flash-tier Neon model — OwnTechnology4271 · 2026-11-19