# Command R+ (04-2024)

> The original 104B Command R+ — Cohere's first flagship RAG and tool-use model with a 128K context window.

- Maker: Cohere
- Line: Command R+
- Canonical page: https://ai-tldr.dev/models/command-r-plus-04-2024/
- Updated: 2026-06-17

## In short

Command R+ (04-2024) is the original release of Cohere's flagship Command R+ model, launched April 4, 2024 as a 104-billion-parameter dense transformer. In the Cohere API it is addressed as command-r-plus (an alias for command-r-plus-04-2024). It has a 128K-token context window and costs $3.00 per million input tokens and $15.00 per million output tokens. You can use it through Cohere and Hugging Face (open weights, research).

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 128K (max output 4K tokens) | [1] |
| Input price | $3.00 / 1M tokens | [2] |
| Output price | $15.00 / 1M tokens | [2] |
| Licence | Proprietary (API); open weights under CC-BY-NC on Hugging Face · open weights | [1] |
| Released | 4 April 2024 | [3] |
| Parameters | 104B (dense) | [1] |
| Knowledge cutoff | February 2023 | [1] |

## Sources

1. [c4ai-command-r-plus model card (Hugging Face)](https://huggingface.co/CohereLabs/c4ai-command-r-plus)
2. [Cohere pricing](https://cohere.com/pricing)
3. [Command R+ docs (official)](https://docs.cohere.com/docs/command-r-plus)

## Overview

Command R+ (04-2024) is the original release of Cohere's flagship Command R+ model, launched April 4, 2024 as a 104-billion-parameter dense transformer. In the Cohere API it is addressed as command-r-plus (an alias for command-r-plus-04-2024). It introduced Cohere's scalable Command R family's high end, targeting real-world enterprise use cases with advanced retrieval-augmented generation (RAG) and multi-step (agentic) tool use.

The model uses a dense autoregressive transformer with Multi-Head Attention (MHA), supports a 128K-token context window, and is optimized for 10 business languages (English, French, Spanish, Italian, German, Brazilian Portuguese, Japanese, Korean, Simplified Chinese, and Arabic), with pre-training spanning 23 languages in total. Its headline features at launch were grounded RAG with inline citations and the ability to combine multiple tools across multiple steps to complete complex tasks.

Cohere also released the weights for research on Hugging Face (CohereLabs/c4ai-command-r-plus) under a CC-BY-NC license. This version has since been superseded: Cohere shipped the faster, cheaper 08-2024 refresh and deprecated command-r-plus-04-2024 on September 15, 2025, recommending migration to command-r-plus-08-2024, command-r-08-2024, or command-a-03-2025.

## Benchmarks

- Open LLM Leaderboard (average): 74.6% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- MMLU (5-shot): 75.7% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- HellaSwag (10-shot): 88.6% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- ARC Challenge (25-shot): 70.99% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- Winogrande (5-shot): 85.4% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- GSM8K (5-shot): 70.7% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))
- TruthfulQA: 56.3% ([source](https://huggingface.co/CohereLabs/c4ai-command-r-plus))

## Pricing

- Input: $3.00 / 1M tokens
- Output: $15.00 / 1M tokens

Source: https://cohere.com/pricing

## Strengths

- First Cohere flagship purpose-built for enterprise RAG with grounded inline citations
- Multi-step tool use combining several tools across multiple steps (agentic workflows)
- Long 128K-token context window for long documents and multi-turn conversations
- Multilingual: optimized for 10 languages, pre-trained on 23
- Open weights published on Hugging Face for research under CC-BY-NC

## Best for

- Enterprise retrieval-augmented generation with citations
- Multi-step tool-using agents
- Multilingual business document Q&A and support
- Long-context document summarization and analysis

## How to access

- [Cohere](https://docs.cohere.com/docs/command-r-plus) — model id `command-r-plus-04-2024`
- [Hugging Face (open weights, research)](https://huggingface.co/CohereLabs/c4ai-command-r-plus) — model id `CohereLabs/c4ai-command-r-plus`

## Command R+ versions

- [Command R+ (08-2024)](https://ai-tldr.dev/models/command-r-plus-08-2024/) (30 August 2024)
- Command R+ (04-2024) (4 April 2024) — this page

## FAQ

### Is Command R+ (04-2024) still available?

No. Cohere deprecated command-r-plus-04-2024 (alias command-r-plus) on September 15, 2025 and recommends migrating to command-r-plus-08-2024, command-r-08-2024, or command-a-03-2025 (described as the strongest-performing model across domains).

### How big is Command R+ (04-2024)?

It is a 104-billion-parameter dense autoregressive transformer using Multi-Head Attention (MHA), with a 128K-token context window and up to 4K output tokens per response.

### How did Command R+ (04-2024) score on benchmarks?

On the Open LLM Leaderboard its model card reports an average of 74.6, including 75.7 on MMLU, 88.6 on HellaSwag, 70.99 on ARC Challenge, 85.4 on Winogrande, 70.7 on GSM8K, and 56.3 on TruthfulQA.

### What did Command R+ (04-2024) cost?

At launch it was priced at $3.00 per million input tokens and $15.00 per million output tokens, per Cohere's pricing page. The later 08-2024 refresh lowered this to $2.50 / $10.00.

## Links

- [Command R+ docs (official)](https://docs.cohere.com/docs/command-r-plus)
- [c4ai-command-r-plus model card (Hugging Face)](https://huggingface.co/CohereLabs/c4ai-command-r-plus)
- [Cohere models overview (context & output limits)](https://docs.cohere.com/docs/models)
- [Cohere deprecations (04-2024 retirement)](https://docs.cohere.com/docs/deprecations)
- [Cohere pricing](https://cohere.com/pricing)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/command-r-plus-04-2024/