Command A 03 2025: benchmark results
Cohere's Command A, a 111B dense language model released in March 2025 under a non-commercial (CC-BY-NC-4.0) research license. Provider: Cohere. Released 2025-03-13. Access: Open.
Aggregate Rating 1573 ± 24, rank #530 of 1635 rated models, from 31 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BALSAM - Factuality | 26.56 | Overall score (0-100, LLM-judged generation and multiple cho | 97.3 |
| BALSAM - Reading Comprehension | 59.8 | Overall score (0-100, LLM-judged generation and multiple cho | 94.4 |
| BALSAM - Text Classification | 39.82 | Overall score (0-100, LLM-judged generation and multiple cho | 87.9 |
| BenchTable - Tech | 79.7 | Weighted Score (%) | 87.5 |
| BALSAM - Text Manipulation | 64.34 | Overall score (0-100, LLM-judged generation and multiple cho | 86.5 |
| BALSAM - Creative Writing | 51.1 | Overall score (0-100, LLM-judged generation and multiple cho | 84.8 |
| BALSAM - Question Answering | 72.2 | Overall score (0-100, LLM-judged generation and multiple cho | 83.8 |
| BALSAM - Translation/Transliteration | 64.95 | Overall score (0-100, LLM-judged generation and multiple cho | 81.1 |
| BALSAM - Information Extraction | 49.92 | Overall score (0-100, LLM-judged generation and multiple cho | 77.1 |
| Vals AI CaseLaw v2 | 64.52 | Accuracy (%) | 76.8 |
| BALSAM - Overall | 55.16 | Mean of category overall scores (0-100) | 75.9 |
| BenchTable - Utility | 62.8 | Weighted Score (%) | 68.9 |
Interactive version: theaggregate.ai/model?slug=command-a-03-2025 · How It Works · Data refreshed daily, snapshot 2026-11-05.