Kimi K2.5 (High): benchmark results
Provider: Moonshot. Released 2026-01-27. Access: Open.
Aggregate Rating 1679 ± 15, rank #419 of 2170 rated models, from 50 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CLBench Life - Self-Tracking Trajectories | 8.9 | Task solving rate (%) on the 45 self-tracking trajectory (fi | 88.9 |
| Grid-Bench T0 pooled exact | 192 | Exact answers (out of 192) | 80 |
| Grid-Bench T1 D2 exact | 56 | Exact answers (out of 64) | 80 |
| Grid-Bench T1 D3 exact | 53 | Exact answers (out of 64) | 80 |
| FutureEval | 8.92 | Unified Forecasting Score | 76.2 |
| SWE-bench Verified | 70.8 | Resolved (%) | 76.1 |
| NonoBench (v1.0) | 43.33 | Standard accuracy (%), correct / non-failed attempts; versio | 75.7 |
| BaFCo - Coarse Layout Analysis (CoT) | 13.61 | Mean average precision at IoU 0.3 (0-100): class-wise averag | 66.7 |
| CLBench Life - Creation & Revision Histories | 13.3 | Task solving rate (%) on the 45 creation and revision histor | 66.7 |
| OneMillion-Bench (CN) | 19.5 | Pass rate (%): share of tasks whose expert score reaches the | 66.7 |
| OneMillion-Bench - Expert Score | 45.1 | Expert score (%): per task, the weighted share of expert-wri | 66.7 |
| Grid-Bench T1 D4 exact | 39 | Exact answers (out of 64) | 60 |
Interactive version: theaggregate.ai/model?slug=kimi-k2-5-high · How It Works · Data refreshed daily, snapshot 2026-11-05.