Open-weight MoE that matched frontier closed models on coding
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 27 models released within six months of it, across 5 benchmarks where scores spread out.
| Benchmark | Score | Rank |
|---|---|---|
MATH Competition-level mathematics problems | 97.4% | #8 / 50 |
Arena Elo Human preference ranking via blind comparisons | 1473 | #16 / 51 |
MMLU Tests knowledge across 57 subjects from STEM to humanities | 89.5% | #23 / 54 |
SWE-bench Real-world GitHub issue resolution | 65.8% | #27 / 39 |
HumanEval Coding ability - generating correct Python functions | 85.7% | #36 / 50 |
TerminalArtificial Analysis Agentic terminal coding tasks requiring multi-step execution | 31.1% | #43 / 58 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 23.8% | #74 / 100 |
GPQA PhD-level science questions even experts struggle with | 75.1% | #75 / 96 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisKimi K2 Thinking | 23.8% |
AA-LCR Artificial AnalysisKimi K2 Thinking | 72% |
CritPt Artificial AnalysisKimi K2 Thinking | 2.6% |
IFBench Artificial AnalysisKimi K2 Thinking | 68.1% |