Model Details
Command R+
Quick answer: Command R+ is Cohere's April 2024 enterprise-focused large language model, designed specifically for RAG (Retrieval-Augmented Generation) and tool use pipelines. It scores 70.7% on GSM8K and 88.6% on HellaSwag. With a 128K context window and CC-BY-NC license, it targets enterprise teams building production RAG systems.
At a Glance
Where Command R+ leads
- Optimised for RAG: built-in grounding and citation capabilities
- 88.6% HellaSwag — strong common-sense reasoning
- 128K context window — fits large document contexts for RAG
- Available on HuggingFace under CC-BY-NC
- Multi-language support (10 languages)
- Purpose-built tool use and function calling for agentic pipelines
Where it lags
- CC-BY-NC license restricts commercial deployment
- 70.7% GSM8K — below frontier models on math reasoning
- Superseded by Command R7B and later models in the Cohere lineup
- March 2024 knowledge cutoff
Best for: Enterprise RAG systems; production document search and Q&A; agentic pipelines requiring grounded generation; teams in the Cohere ecosystem.
What Command R+ Is
Command R+ (released April 4, 2024) was Cohere's flagship model designed for enterprise RAG and agentic deployments. Unlike models optimised for benchmark performance, Command R+ was explicitly designed for retrieval-augmented workflows — it was trained with grounding, citation, and multi-step tool use as first-class capabilities.
The model supports 10 languages (English, French, Spanish, Italian, German, Brazilian Portuguese, Japanese, Korean, Arabic, Chinese) and was intended for enterprise deployments requiring robust multilingual document processing.
Its CC-BY-NC license means free use for research and non-commercial applications, but commercial deployments require a Cohere Enterprise agreement. For commercial open-weight RAG pipelines, newer MIT-licensed alternatives (e.g., Llama 3.1 70B, DeepSeek-V3) may be preferable depending on use case.
Specifications
| Field | Value |
|---|---|
| Organization | Cohere |
| Context window | 128,000 tokens |
| License | CC-BY-NC |
| HuggingFace | CohereForAI/c4ai-command-r-plus |
| Release date | April 4, 2024 |
| Knowledge cutoff | March 2024 |
| Modality | Text only |
| Languages | 10 (EN, FR, ES, IT, DE, PT, JA, KO, AR, ZH) |
Pricing
Command R+ is available as open weights under CC-BY-NC on Hugging Face. Commercial API access via the Cohere platform.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| GSM8K | 70.7% | Benchgen evaluation | 2025-07 |
| HellaSwag | 88.6% | Benchgen evaluation | 2025-07 |
Command R+ vs Alternatives
| Model | GSM8K | HellaSwag | License | RAG focus |
|---|---|---|---|---|
| Command R+ (Cohere) | 70.7% | 88.6% | CC-BY-NC | Yes |
| Llama 3.1 70B Instruct | — | — | Llama 3.1 | General |
| DeepSeek-V3 | — | — | MIT | General |
| Qwen2 72B Instruct | 89.5% | — | Qianwen | General |
For pure RAG pipelines within the Cohere ecosystem, Command R+ provides first-class grounding and citation. For general-purpose open-weight deployments with permissive licensing, Llama 3.1 70B (Llama 3.1 License) or DeepSeek-V3 (MIT) are preferred.
Related Models & Benchmarks
- Compare: Llama 3.1 70B Instruct · Qwen2 72B Instruct · DeepSeek-V3
- Benchmarks: GSM8K · HellaSwag
Frequently Asked Questions
What is Command R+?
Command R+ is Cohere's April 2024 enterprise model designed for RAG, scoring 70.7% GSM8K and 88.6% HellaSwag with a 128K context window. It excels in grounded generation and multilingual document processing.Is Command R+ open source?
The weights are available under CC-BY-NC on Hugging Face. Commercial use requires a Cohere Enterprise license — CC-BY-NC prohibits commercial deployment without permission.What makes Command R+ different from other models?
Command R+ was specifically designed for RAG pipelines with built-in grounding and citation capabilities. Most general-purpose models treat RAG as an application pattern; Command R+ treats it as a core model capability.Specs from Cohere's official Command R+ release (April 2024) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.