Przemek Chojecki | PC(@prz_chojecki):Muse Spark 1.2 is better than GPT-5.5 xhigh and only slightly worse than Kimi K3 on our ErdosBench. We've tested the new model from Meta on 226 research-level math problems and it solved 40 / 226 problems and gave many interesting partial solutions. Muse Spark 1.2 is a strong entrant: good proof hygiene, high B-grade review yield, no rejected strong claims, but fewer decisive A-grade closures.

Przemek Chojecki | PC

@prz_chojecki

Reasoning Data + Evals for LLMs @ PhD in mathematics.

Joined April 2016

1.1K Following 14.7K Followers

Przemek Chojecki | PC@prz_chojecki

2026.08.06 19:16

Muse Spark 1.2 is better than GPT-5.5 xhigh and only slightly worse than Kimi K3 on our ErdosBench. We've tested the new model from Meta on 226 research-level math problems and it solved 40 / 226 problems and gave many interesting partial solutions. Muse Spark 1.2 is a strong entrant: good proof hygiene, high B-grade review yield, no rejected strong claims, but fewer decisive A-grade closures.