Command: benchmark results
Cohere's Command, the company's flagship generative and conversational model, released in beta through a retrieval-augmented-generation chat API in September 2023. Provider: Cohere. Released 2023-09-29. Access: API.
Aggregate Rating 1371 ± 50, rank #1331 of 1656 rated models, from 30 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM NaturalQuestions (Open) | 77.69 | F1 (%) | 88.9 |
| HELM NarrativeQA | 74.88 | F1 (%) | 60 |
| HELM NaturalQuestions (Closed) | 39.11 | F1 (%) | 60 |
| HELM Lite - LegalBench - Abercrombie | 55.79 | Exact Match (%) | 45 |
| HELM Lite - LegalBench - PROA | 88.42 | Exact Match (%) | 42.8 |
| HELM Lite - LegalBench - Function of Decision Section | 36.51 | Exact Match (%) | 40 |
| HELM Lite - LegalBench | 57.75 | Exact Match (%) | 33.3 |
| HELM Lite | 35.79 | Mean win rate (imported) | 30.3 |
| HELM (Stanford) | 32.68 | Mean Win Rate (%) | 26.7 |
| HELM Lite - MATH Level 1 - Geometry | 26.32 | Equivalent (CoT) (%) | 23.3 |
| HELM Lite - LegalBench - Corporate Lobbying | 67.14 | Exact Match (%) | 22.2 |
| HELM Lite - MATH Level 1 - Intermediate Algebra | 13.46 | Equivalent (CoT) (%) | 18.3 |
Interactive version: theaggregate.ai/model?slug=command · How It Works · Data refreshed daily, snapshot 2026-10-20.