Kimi K2.6: benchmark results
Moonshot Kimi K2.6 open 1T-parameter MoE model. Provider: Moonshot. Released 2026-04-20. Access: Open.
Aggregate Rating 1759 ± 3, rank #104 of 1656 rated models, from 1741 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| APIFlow-Bench 1.0 - 5-Subtask Chain | 100 | Pass Rate (%) | 100 |
| APIFlow-Bench 1.0 - Schema | 93.2 | Pass Rate (%) | 100 |
| CESBench - Countermeasures | 85.5 | Composite score (%; mean over the four task types of the ite | 100 |
| CESBench - Evaluation | 80.5 | Composite score (%; mean over the four task types of the ite | 100 |
| CLIR-Bench - Intervention Decision | 91.67 | Accuracy (%) on immediate intervention decision, 4-option mu | 100 |
| CLIR-Bench - Intervention Response | 93.33 | Accuracy (%) on intervention response, 4-option multiple-cho | 100 |
| CLIR-Bench - Missingness Awareness | 73.33 | Accuracy (%) on missingness awareness (whether the observati | 100 |
| CircuitReason-1k - AC and Phasor | 84.09 | Accuracy (%; exact problem-level correctness on the 308 AC a | 100 |
| CircuitReason-1k - Long-Horizon Reasoning | 83.73 | Accuracy (%; exact problem-level correctness on the 332 long | 100 |
| CircuitReason-1k - Transient and Frequency | 87.86 | Accuracy (%; exact problem-level correctness on the 173 tran | 100 |
| DeskCraft (Standard) | 33.8 | Task success rate (%) on the 386-task standard split, verifi | 100 |
| FIKA-Bench | 25.1 | Overall strict answer accuracy (%) on FIKA-Bench, open-ended | 100 |
Interactive version: theaggregate.ai/model?slug=kimi-k2-6 · How It Works · Data refreshed daily, snapshot 2026-10-20.