Open-weight native-vision flagship with 2.8T parameters and a one-million-token context window.
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 76 models released within six months of it, across 3 benchmarks where scores spread out.
| Benchmark | Score | Rank |
|---|---|---|
AA Coding (Historical) Archived release-era coding composite, excluded from current rankings because comparable current-series coverage is unavailable | 76.2% | #8 / 22 |
GPQAArtificial Analysis PhD-level science questions even experts struggle with | 93.5% | #10 / 96 |
AA Intelligence (Historical) Archived release-era composite, not comparable to the current versioned Intelligence Index used for rankings | 57.1% | #10 / 22 |
MMMUvals.ai College-level multimodal reasoning across 30+ disciplines | 88.2% | #11 / 51 |
aaAgentic | 50.1% | #11 / 16 |
LiveCodeBenchvals.ai Contamination-free competitive programming (filtered by cutoff date) | 87.2% | #16 / 62 |
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 46.9% | #20 / 102 |
MMLU-Provals.ai Harder 10-option successor to MMLU; more reasoning-focused | 88% | #22 / 61 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisKimi K3 (Max) | 46.9% |
MMMU-Pro Artificial AnalysisKimi K3 (Max) | 80.5% |
Terminal-Bench 2.1 Artificial AnalysisKimi K3 (Max) | 85% |
Terminal-Bench 4.0 Artificial AnalysisKimi K3 (Max) | 12.6% |
SciCode Artificial AnalysisKimi K3 (Max) | 59.5% |
AA-LCR Artificial AnalysisKimi K3 (Max) | 88.7% |
CritPt Artificial AnalysisKimi K3 (Max) | 23.4% |
ITBench SRE Artificial AnalysisKimi K3 (Max) | 47.7% |