MiniMax M2.5 (Thinking): benchmark results
Provider: MiniMax. Released 2026-02-12. Access: Open.
Aggregate Rating 1565 ± 19, rank #814 of 2151 rated models, from 27 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Grid-Bench T1 D4 exact | 44 | Exact answers (out of 64) | 80 |
| Grid-Bench T1 pooled exact | 150 | Exact answers (out of 192) | 80 |
| LisanBench | 0.06 | Mean Path Length / Current Maximum | 64.1 |
| Grid-Bench T1 D2 exact | 55 | Exact answers (out of 64) | 60 |
| Grid-Bench T16 D3 exact | 5 | Exact answers (out of 64) | 50 |
| Grid-Bench T64 pooled exact | 1 | Exact answers (out of 192) | 50 |
| Grid-Bench T1 D3 exact | 51 | Exact answers (out of 64) | 40 |
| Grid-Bench T16 pooled exact | 11 | Exact answers (out of 192) | 40 |
| Grid-Bench T32 pooled exact | 5 | Exact answers (out of 192) | 40 |
| Grid-Bench T8 pooled exact | 31 | Exact answers (out of 192) | 40 |
| Grid-Bench T16 D2 exact | 6 | Exact answers (out of 64) | 30 |
| Grid-Bench T16 D4 exact | 0 | Exact answers (out of 64) | 30 |
Interactive version: theaggregate.ai/model?slug=minimax-m2-5-thinking · How It Works · Data refreshed daily, snapshot 2026-10-17.