You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

Model Card for Cohere Labs Command R+ 08-2024

Model Summary

Cohere Labs Command R+ 08-2024 is an open weights research release of a 104B billion parameter model with highly advanced capabilities, this includes Retrieval Augmented Generation (RAG) and tool use to automate sophisticated tasks. The tool use in this model generation enables multi-step tool use which allows the model to combine multiple tools over multiple steps to accomplish difficult tasks. Command R+ 08-2024 is a multilingual model trained on 23 languages and evaluated in 10 languages. Command R+ 08-2024 is optimized for a variety of use cases including reasoning, summarization, and question answering.

Cohere Labs Command R+ 08-2024 is part of a family of open weight releases from Cohere Labs and Cohere. Our smaller companion model is Cohere Labs Command R 08-2024.

• Point of Contact: Cohere Labs
• License:CC-BY-NC, requires also adhering to Cohere Lab's Acceptable Use Policy
• Model: coherelabs-command-r-plus-08-2024
• Model Size: 104 billion parameters
• Context length: 128K

Try Cohere Labs Command R+

You can try out Cohere Labs Command R+ before downloading the weights in our hosted Hugging Face Space.

Usage

Please use transformers version 4.39.1 or higher

# pip install 'transformers>=4.39.1'
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "CohereLabs/c4ai-command-r-plus-08-2024"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# Format message with the command-r-plus-08-2024 chat template
messages = [{"role": "user", "content": "Hello, how are you?"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>

gen_tokens = model.generate(
 input_ids, 
 max_new_tokens=100, 
 do_sample=True, 
 temperature=0.3,
 )

gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)

Model Details

Input: Models input text only.

Output: Models generate text only.

Model Architecture: This is an auto-regressive language model that uses an optimized transformer architecture. After pretraining, this model uses supervised fine-tuning (SFT) and preference training to align model behavior to human preferences for helpfulness and safety. We use grouped query attention (GQA) to improve inference speed.

Languages covered: The model has been trained on 23 languages (English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic, Simplified Chinese, Russian, Polish, Turkish, Vietnamese, Dutch, Czech, Indonesian, Ukrainian, Romanian, Greek, Hindi, Hebrew, and Persian) and evaluated on 10 languages (English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic, Simplified Chinese).

Context length: Command R+ 08-2024 supports a context length of 128K.

Grounded Generation and RAG Capabilities:

Command R+ 08-2024 has been specifically trained with grounded generation capabilities. This means that it can generate responses based on a list of supplied document snippets, and it will include grounding spans (citations) in its response indicating the source of the information. This can be used to enable behaviors such as grounded summarization and the final step of Retrieval Augmented Generation (RAG). This behavior has been trained into the model via a mixture of supervised fine-tuning and preference fine-tuning, using a specific prompt template. Deviating from this prompt template may reduce performance, but we encourage experimentation.

Command R+ 08-2024’s grounded generation behavior takes a conversation as input (with an optional user-supplied system preamble, indicating task, context and desired output style), along with a list of retrieved document snippets. The document snippets should be chunks, rather than long documents, typically around 100-400 words per chunk. Document snippets consist of key-value pairs. The keys should be short descriptive strings, the values can be text or semi-structured.

By default, Command R+ 08-2024 will generate grounded responses by first predicting which documents are relevant, then predicting which ones it will cite, then generating an answer. Finally, it will then insert grounding spans into the answer. See below for an example. This is referred to as accurate grounded generation.

The model is trained with a number of other answering modes, which can be selected by prompt changes. A fast citation mode is supported in the tokenizer, which will directly generate an answer with grounding spans in it, without first writing the answer out in full. This sacrifices some grounding accuracy in favor of generating fewer tokens.

Comprehensive documentation for working with Command R+ 08-2024's grounded generation prompt template can be found here, here and here.

You can render the Grounded Generation prompt template by using the function apply_grounded_generation_template(). The code snippet below shows a minimal working example on how to render this prompt.

Single-Step Tool Use Capabilities ("Function Calling"):

Single-step tool use (or “Function Calling”) allows Command R+ 08-2024 to interact with external tools like APIs, databases, or search engines. Single-step tool use is made of two model inferences:

• Tool Selection: The model decides which tools to call and with what parameters. It’s then up to the developer to execute these tool calls and obtain tool results.
• Response Generation: The model generates the final response given the tool results. You can learn more about single step tool use in our documentation.

Command R+ 08-2024 has been specifically trained with single-step tool use (or “Function Calling”) capabilities. These have been trained into the model via a mixture of supervised fine-tuning and preference fine-tuning, using a specific prompt template. Deviating from this prompt template may reduce performance. This is why we recommend using the prompt template described below.

Command R+ 08-2024’s single-step tool use functionality takes a conversation as input (with an optional user-system preamble), along with a list of available tools. The model will then generate a json-formatted list of actions to execute on a subset of those tools. Command R+ 08-2024 may use one of its supplied tools more than once.

The model has been trained to recognise a special directly_answer tool, which it uses to indicate that it doesn’t want to use any of its other tools. The ability to abstain from calling a specific tool can be useful in a range of situations, such as greeting a user, or asking clarifying questions. We recommend including the directly_answer tool, but it can be removed or renamed if required.

Comprehensive documentation for working with Command R+ 08-2024's single-step tool use prompt template can be found here and here.

You can render the single-step tool use prompt template by using the function apply_tool_use_template(). The code snippet below shows a minimal working example on how to render this prompt.

Command R+ 08-2024 also supports Hugging Face's tool use API to render the same prompt.

Multi-Step Tool Use Capabilities ("Agents"):

Multi-step tool use is suited for building agents that can plan and execute a sequence of actions using multiple tools. Unlike single-step tool use, the model can perform several inference cycles, iterating through Action → Observation → Reflection until it decides on a final response. For more details, refer to our documentation on multi-step tool use.

Command R+ 08-2024 has been specifically trained with multi-step tool use (or “Agents”) capabilities. These have been trained into the model via a mixture of supervised fine-tuning and preference fine-tuning, using a specific prompt template. Deviating from this prompt template may reduce performance. This is why we recommend using the prompt template described below.

The prompt template is not yet available in HuggingFace. However, comprehensive documentation for working with Command R+ 08-2024's multi-step tool use prompt template can be found here and here.

Code Capabilities:

Command R+ 08-2024 has been optimized to interact with your code, by requesting code snippets, code explanations, or code rewrites. It might not perform well out-of-the-box for pure code completion. For better performance, we also recommend using a low temperature (and even greedy decoding) for code-generation related instructions.

Model Card Contact

For errors or additional questions about details in this model card, contact labs@cohere.com

Terms of Use:

We hope that the release of this model will make community-based research efforts more accessible, by releasing the weights of a highly performant 104 billion parameter model to researchers all over the world. This model is governed by a CC-BY-NC, requires also adhering to Cohere Lab's Acceptable Use Policy

Try Chat:

You can try Command R+ 08-2024 chat in the playground here. You can also use it in our dedicated Hugging Face Space here.

Cite

To cite this model, use:

@misc {cohere_for_ai_2024,
 author = { {Cohere Labs} },
 title = { c4ai-command-r-plus-08-2024 },
 year = 2024,
 url = { https://huggingface.co/CohereLabs/c4ai-command-r-plus-08-2024 },
 doi = { 10.57967/hf/3135 },
 publisher = { Hugging Face }
}

Latest Cohere Labs Command models • 12 items • Updated Nov 6, 2025 • 34