OpenAI's GPT-6 Astra cracked 10 open math problems and improved a 1970s sphere-packing bound
morqon · x · 2026-09-08
a16z's Lisha Li sits down with OpenAI researchers Mark Sellke and Mehtaab Sawhney on AI and mathematics, recorded before GPT-6 Astra's launch.
- It started when Sawhney pasted GPT-5 an Erdős problem still listed as open; five minutes later the model surfaced a paper that had solved it, eventually uncovering published solutions to 10 more problems thought to be open.
- Astra went further: told to "go have fun" with a high-dimensional sphere-packing problem, it improved a bound that had stood since the 1970s.
- Context: an extraordinary few weeks for AI math, including progress around the resolution of Navier–Stokes, the twin prime conjecture, and ten open problems from sphere packing to non-sofic groups.
- The conversation covers how Astra found these results and how much of a mathematician's job it can now do.
Related event: OpenAI's Astra Cracks Progress on Ten Open Math Problems(5 posts)→
More from Models
- Blender head-to-head: same prompt, 12 seconds, and one frontier model is in a different class — ZeroStateReflex · 2026-09-09
- Netlify adds GPT-6 Astra, Gemini 3.8 Flash, Claude Fable 5.1 and smarter Agent Runner scoping — thisiskp_ · 2026-09-09
- GLM-5.3-Flash tops agentic tool-call leaderboard at 78%, priced at just $0.50/M output tokens — shensi · 2026-09-09
- Artificial Analysis Updated Benchmarks Twice in 4 Days for Astra — py-net · 2026-09-09
- Gary Marcus: it was the harness, not the model — hidden risk of opaque scaffolding — GaryMarcus · 2026-09-09
- Inception ships Mercury 2.5: most capable diffusion LLM at 1,107 tokens/sec, 260K context — StefanoErmon · 2026-09-09