Successor to K2.5 with stronger reasoning and Arena Elo; trades some agentic score for benchmark gains
SI v3 uses weighted percentages from one Artificial Analysis configuration. Ranking requires all five benchmarks. Partial scores average only available weights. Scores are not comparable to SI v2 or a probability of reaching singularity.
Compared with models released within six months of it.
| Benchmark | Score | Rank |
|---|---|---|
Arena EloArtificial Analysis Human preference ranking via blind comparisons | 1483 | #14 / 51 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 86.8% | #18 / 62 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 86.3% | #19 / 51 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 43.9% | #24 / 58 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 87.6% | #25 / 61 |
GPQAArtificial Analysis PhD-level science questions even experts struggle with | 91.1% | #36 / 95 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 37.5% | #46 / 95 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisKimi K2.6 (Reasoning) | 37.5% |
MMMU-Pro Artificial AnalysisKimi K2.6 (Reasoning) | 79.4% |
Terminal-Bench 2.1 Artificial AnalysisKimi K2.6 (Reasoning) | 65.9% |
Terminal-Bench 4.0 Artificial AnalysisKimi K2.6 (Reasoning) | 0.5% |
SciCode Artificial AnalysisKimi K2.6 (Reasoning) | 51.5% |
AA-LCR Artificial AnalysisKimi K2.6 (Reasoning) | 81% |
CritPt Artificial AnalysisKimi K2.6 (Reasoning) | 8% |
ITBench SRE Artificial AnalysisKimi K2.6 (Reasoning) | 31.2% |
IFBench Artificial AnalysisKimi K2.6 (Reasoning) | 76% |