OpenAI has largely automated training experimental models; new model cracks 100 open math problems
kimmonismus · x · 2026-09-22
Via The Information, OpenAI has 'largely automated the process of training new experimental models' internally — widely read as a real step toward recursive self-improvement. OpenAI also blogged that a newly trained model solved 100 open long-standing math problems, surprising its own researchers. Elsewhere, Xiaomi's MiMo shipped an impressive model while Grok 4.7 underperformed. (Third-party reporting; details unverified.)
Related event: The Information: OpenAI has largely automated experimental model training(7 posts)→
More from Models
- Rumors swirl of OpenAI unveiling a Grok bot rival at DevDay — imjustnewatai · 2026-09-22
- New Local-LLM Primitive: "Span" Answers Point to Snippets Instead of Enumerating — generativist · 2026-09-22
- Grok 4.7 lands in Devin at 59.4% on FrontierCode, but Xiaomi MiMo-V2.6 undercuts it — airesearch12 · 2026-09-22
- Xiaomi's model jumps near the frontier with just 30 RL steps, no plateau in sight — teortaxesTex · 2026-09-22
- Xiaomi's MiMo-V2.6-Pro-RL Climbs Hugging Face Trending Charts — XiaomiMiMo · 2026-09-22
- LLMs lack artistic intent, and writing may resist outcome-based RL — teortaxesTex · 2026-09-22