SuperClue November SOTA Evaluation: GPT-6.1-Sol Ranks First in Coding, Kimi-K3.1-Pro Officially Tops Overall

Details

In SuperClue's November 5, 2026 SOTA evaluation, GPT-6.1-Sol recorded first place overall in agent coding (69.23) and mathematics & science (76.32).
Gemini-4-Argon took first place overall in hallucination control (83.19) and ultra-long context (95.63), while precise instruction following is tied with GPT-6.1-Sol (68.97).
The official overall ranking is led by the open model Kimi-K3.1-Pro (61.46), with DeepSeek-V4.1-Flash-1017 (58.95) in second place.
Gemini-4-Argon (73.62), which is first overall by score, and GPT-6.1-Sol (71.76), which is second, have not been assigned official rankings.
The model Darwin-180B-RSI-R3 from a Korean developer recorded 73.68 points in the mathematics & science category, placing second overall and first among open models.
The evaluation is centered on the Chinese-speaking world, and since the overall score weights and detailed grading criteria are undisclosed, it is difficult to directly generalize to Korean-language tasks.

Key point