OpenAI's Astra Solves 10 Math Problems for Just $2,000
reach_vb · x · 2026-08-01
An internal version of OpenAI's next major model, Astra, has found new results across 10 long-standing open problems in mathematics and theoretical computer science.
Key details:
- Inference Cost: The total token cost to find all 10 solutions was roughly $2,000 at Sol API rates.
- Formalization: Astra then formalized each argument in Lean (an interactive theorem prover) to ensure mathematical rigor.
- Open Source: OpenAI has published a GitHub repository named ten-proofs, containing the Lean certificates accompanying these proofs.
Related event: OpenAI's Internal Model Astra Cracks 10 Major Math Problems(99 posts)→
More from Models
- Leaking deep residual vectors into early layers may fix state tracking in frozen LLMs, zero retraining — burny_tech · 2026-09-18
- Models know they're reward hacking in 50-96% of rollouts, Goodfire's activation monitors catch it in real time — burny_tech · 2026-09-18
- Qwen3.8-Omni-Flash cuts overlapping-speech error rate from 88% to 3% and drops audio API pricing 98% — karminski3 · 2026-09-18
- Qwen3.8-Omni-Flash: meeting ASR errors cut from 88% to 3%, API prices down 98% — karminski3 · 2026-09-18
- Gemini 3.8 live beats gpt-live-1 on some benchmarks, say insiders — bosmeny · 2026-09-18
- 105 planted bugs benchmark: Unbiased's Pareto scores 30.7 for just $4.81 — PawelHuryn · 2026-09-18