Command A: benchmark results
Cohere Command A can attach citations to supplied documents and tool results, supporting answers traceable to enterprise data. Downloadable weights carry a non-commercial license. Provider: Cohere. Released 2025-03-13. Access: Open.
Aggregate Rating 1581 ± 10, rank #508 of 1656 rated models, from 361 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Evals for Every Language - Chinese (zh) | 75.55 | Average Score (%) | 95.5 |
| ProLLM - OpenBook Q&A | 86.3 | Score (%) | 92.4 |
| ProLLM - Summarization | 91.2 | Score (%) | 89.7 |
| Evals for Every Language - Portuguese (pt) | 77.5 | Average Score (%) | 87.7 |
| Evals for Every Language - Indonesian (id) | 74.33 | Average Score (%) | 83.3 |
| Evals for Every Language - Awadhi (awa) | 71.15 | Average Score (%) | 83.1 |
| Evals for Every Language - Classification | 95.36 | Average Score (%) | 82.4 |
| AA Omniscience - Software Engineering (SWE) - Dart | 28 | Accuracy (%) | 82.1 |
| Evals for Every Language - Galician (gl) | 73.38 | Average Score (%) | 80.3 |
| Evals for Every Language - Persian (fa) | 73.12 | Average Score (%) | 80 |
| ProLLM - SQL Disambiguation | 40.6 | Score (%) | 76.5 |
| Vals AI CaseLaw v2 | 64.52 | Accuracy (%) | 75.4 |
Interactive version: theaggregate.ai/model?slug=command-a · How It Works · Data refreshed daily, snapshot 2026-10-20.