Side-by-side benchmark comparison across knowledge, coding, math, and reasoning.
DeepSeek V3.1 (Reasoning) wins overall with a score of 25 vs 23 (2 point difference).DeepSeek V3.1 (Reasoning) wins 4 out of 4 categories.
DeepSeek V3.1 (Reasoning)
31.8
Kimi K2
29.8
DeepSeek V3.1 (Reasoning)
26
Kimi K2
24
DeepSeek V3.1 (Reasoning)
33
Kimi K2
31
DeepSeek V3.1 (Reasoning)
31
Kimi K2
29
DeepSeek V3.1 (Reasoning) scores higher overall with 25 vs 23, a difference of 2 points across all benchmarks.
DeepSeek V3.1 (Reasoning) leads in knowledge tasks with an average score of 31.8 vs 29.8.
DeepSeek V3.1 (Reasoning) leads in coding with an average score of 26 vs 24.
DeepSeek V3.1 (Reasoning) leads in math with an average score of 33 vs 31.
DeepSeek V3.1 (Reasoning) leads in reasoning with an average score of 31 vs 29.