minimax-01: benchmark results
MiniMax's MiniMax-01 (MiniMax-Text-01), a 456B-parameter (45.9B active) mixture-of-experts model using a Lightning Attention mechanism, with a context window up to 4 million tokens. Provider: MiniMax. Released 2025-01-14. Access: API.
Unified ELO 1557 ± 18, rank #589 of 1639 rated models, from 84 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGC-Bench - moh_x | 0.82 | Dataset z-score | 92.6 |
| AGC-Bench - ocw | 1.33 | Dataset z-score | 87.8 |
| AGC-Bench - story_quality | 1 | Dataset z-score | 86.4 |
| AGC-Bench - science_analogies | 0.55 | Dataset z-score | 80 |
| AGC-Bench - fann_or_flop | 0.72 | Dataset z-score | 79.3 |
| Context Arena MRCR (2-needle) | 31.4 | AUC@1M (%) | 78.8 |
| AGC-Bench - unfun_corpus | 0.28 | Dataset z-score | 78.2 |
| DuckDB-NSQL | 61.3 | Execution Accuracy (%) | 74.5 |
| AGC-Bench - mops | 0.7 | Dataset z-score | 74.4 |
| AGC-Bench - speak_to_structure | 0.63 | Dataset z-score | 72.8 |
| AGC-Bench - fig_qa | -0.14 | Dataset z-score | 70.7 |
| AGC-Bench - creatset | 0.66 | Dataset z-score | 68.3 |
Interactive version: theaggregate.ai/model?slug=minimax-01 · How It Works · Data refreshed daily, snapshot 2026-10-09.