OpenAI internal model reportedly solves a unit-distance problem 48% of the time with enough compute
iruletheworldmo · x · 2026-07-21
- A quoted post discusses an unreleased OpenAI internal model that reportedly solves the unit distance problem 48% of the time when given enough compute, fully autonomously and in one shot, without Lean or any special harness.
- The author argues the graph likely reflects a substantial compute investment, because each plotted point represents 100 attempts and the inferred per-attempt cost may be in the $50K range at the high end.
- The broader implication is that a relatively modest amount of compute can already buy a nontrivial chance of getting a correct solution from the model, even before any external proof tooling is added.
Related event: Report: Unreleased OpenAI Model Solves Math Problem for $100k(3 posts)→
More from Research
- Google DeepMind pilots world's first double-blind evaluations for frontier models — scychan_brains · 2026-10-26
- Rewriting scaling law study: optimal synthetic data share rises from 10% to 32% with compute — heghbalz · 2026-10-26
- FLASH: GPU-native deformable sim learns cloth folding in minutes on one RTX 5090 — ChongZzZhang · 2026-10-26
- LEMON: zero-shot LLM misbehavior monitor flags hallucinations from attention maps — yoavartzi · 2026-10-26
- Behind the EMNLP keynote: LMs as simulated learners to generate child language acquisition hypotheses — najoungkim · 2026-10-26
- The paper behind the EMNLP 2026 keynote on bigram 'simulated learners' — najoungkim · 2026-10-26