Chinese labs have not focused on frontier Math in its LLM development but I've been told Kimi K3 was the 1st Chinese model that was usable for this purpose, so not a surprise to see 3.1 improve on tha

What happened: A post said Chinese labs have not focused on frontier Math in LLM development, but Kimi K3 was the first Chinese model usable for that purpose, so Kimi 3.1 improving on it was not a surprise. Key numbers: Kimi K3.1 Pro enters MathArena as Why it matters: The post said the author had suspected Chinese labs might be more than 6 months behind, and suggested DS could improve faster if it works on frontier math data. It said the 0731 release showed focus on usability for work and SWE applications via the WorkBuddy partnership, with improved science and math as logical next steps.

A post said Chinese labs have not focused on frontier Math in LLM development, but Kimi K3 was the first Chinese model usable for that purpose, so Kimi 3.1 improving on it was not a surprise.

The post said the author had suspected Chinese labs might be more than 6 months behind, and suggested DS could improve faster if it works on frontier math data. It said the 0731 release showed focus on usability for work and SWE applications via the WorkBuddy partnership, with improved science and math as logical next steps.