CohereLabs / command-a-plus-05-2026-w4a4

Model Card for Command A+

Model Summary

Command A+ is an open source model with 25 billion active parameters and 218B total parameters model optimized for agentic, multilingual, and reasoning-heavy tasks with a focus on enterprise performance, while also providing support for vision inputs for processing image inputs.

Developed by: Cohere and Cohere Labs

• Point of Contact: Cohere Labs
• License: Apache 2.0
• Model: command-a-plus-05-2026
• Model Size: 25B active parameters, 218B total parameters
• Context length: 128K input

For more details about this model, please check out our blog post .

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

Available quantizations

The following quantizations are available with example minimum GPU requirements

All three quantizations show negligible differences in benchmark quality and performance. Our recommended quantization for most uses is W4A4 which boasts superior speed and latency characteristics alongside a smaller hardware footprint.

For more details, please check out our blog post .

Usage

Transformers

Please install transformers from the source repository that includes the necessary changes for this model.

# pip install transformers
from transformers import AutoTokenizer, AutoModelForImageTextToText

model_id = "CohereLabs/command-a-plus-05-2026-w4a4"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(model_id)

# Format message with the command-a-plus-05-2026-w4a4 chat template
messages = [{"role": "user", "content": "What has keys but can't open locks?"}]
input_ids = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt",
)

gen_tokens = model.generate(
    input_ids,
    max_new_tokens=4096,
    do_sample=True,
    temperature=0.6,
    top_p=0.95
)

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

As a result, you should get an output that looks like this, where the thinking is generated between the <START_THINKING> and <END_THINKING> :

<|START_THINKING|>The user asks a riddle: "What has keys but can't open locks?" The answer is a piano (or keyboard). So respond with answer.<|END_THINKING|>

You can also use the model directly using transformers pipeline abstraction:

from transformers import pipeline
import torch

model_id = "CohereLabs/command-a-plus-05-2026-w4a4"
tokenizer = AutoTokenizer.from_pretrained(model_id)

pipe = pipeline(
    "text-generation",
    model=model_id,
    dtype="auto",
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Explain the Transformer architecture"},
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)

outputs = pipe(
    messages,
    max_new_tokens=300,
)
print(outputs[0]["generated_text"][-1])

vLLM -- Special instructions for W4A4

Command A+ w4a4 can only run on vLLM >=0.21.0 . W4A4 and accurate response parsing also requires installing Cohere’s melody library.

uv pip install vllm>=0.21.0
uv pip install transformers
uv pip install cohere_melody>=0.9.0

Then the vllm server can be started with the following command:

# This is for B200, adjust tp for your device
vllm serve CohereLabs/command-a-plus-05-2026-w4a4 -tp 1 --tool-call-parser cohere_command4 --reasoning-parser cohere_command4 --enable-auto-tool-choice

We recommend using the following set of sampling parameters for generation: temperature=0.9 , top_p=0.95 , repetition_penalty=1.04 .

Quantization Methodology: Reasoning models pay an outsized quantization tax: long decoding traces compound per-token errors, so naive low-bit conversion typically shows up as visible regressions on hard benchmarks. To mitigate this, we quantize selectively and use distillation to close the residual quality gap. We apply NVFP4 W4A4 quantization (4-bit weights and activations, with two-level scaling) to the MoE experts only. The attention path, i.e., Q/K/V/O projections, the KV cache, and attention compute, is kept at full precision. MoE experts dominate total parameter count, so quantizing them to 4 bits brings the model within the memory budget of a single B200 and accelerates the expert GEMMs that bottleneck short-to-medium-context decode. Furthermore, we use Quantization-Aware Distillation (QAD) in the post-training phase: the quantized student is trained to match the full-precision teacher's output distribution, with fake quantization operators in the forward pass and straight-through estimators on the backward.

Model Details

Input : Text and images.

Output : Model generates text.

Model Architecture : Command A+ is a decoder-only Sparse Mixture-of-Experts Transformer Model. With 25B active parameters and 218B total parameters, it has 128 experts, out of which 8 are active per token, and a single shared expert is applied to all tokens. The attention layers interleave sliding-window attention layers with Rotational Positional Embeddings and global attention layers without positional embeddings in a 3:1 ratio, as first introduced in Command A. The sparse MoE layer is trained in a fully dropless manner and uses a token-choice router. We use additive-bias-based load balancing to encourage balanced token load across all experts, and swap out the softmax router activation function with a normalized sigmoid over the topk expert logits per token.

Languages covered: The model has been trained on 48 languages: English, Arabic, Bulgarian, Bengali, Catalan, Czech, Danish, German, Greek, Spanish, Estonian, Persian, Finnish, Filipino, French, Irish, Hebrew, Hindi, Croatian, Hungarian, Indonesian, Icelandic, Italian, Japanese, Korean, Lithuanian, Latvian, Malay, Maltese, Dutch, Norwegian, Punjabi, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Serbian, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Chinese.

Context Length: Command A+ supports a context length of 128K & 64K output length.

Tool Use Capabilities:

Command A+ has been specifically trained with conversational tool use capabilities. This allows the model to interact with external tools like APIs, databases, or search engines.

Tool use with Command A+ is supported through chat templates in Transformers. We recommend providing tool descriptions using JSON schema.

from transformers import AutoTokenizer

model_id = "CohereLabs/command-a-plus-05-2026-w4a4"
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Define tools
tools = [{
    "type": "function",
    "function": {
        "name": "query_daily_sales_report",
        "description": "Connects to a database to retrieve overall sales volumes and sales information for a given day.",
        "parameters": {
            "type": "object",
            "properties": {
                "day": {
                    "description": "Retrieves sales data for this day, formatted as YYYY-MM-DD.",
                    "type": "string",
                }
            },
            "required": ["day"],
        },
    },
}]

# Define conversation input
conversation = [
    {"role": "user", "content": "Can you provide a sales summary for 29th September 2023?"}
]

# Tokenize the Tool Use prompt directly
input_ids = tokenizer.apply_chat_template(
    conversation=conversation,
    tools=tools,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt",
)

tool_call = {"name": "query_daily_sales_report", "arguments": {"day": "2023-09-29"}}
thinking = "I will use the query_daily_sales_report tool to find the sales summary for 29th September 2023."
conversation.append({"role": "assistant", "tool_calls": [{"id": "0", "type": "function", "function": tool_call}], "thinking": thinking})

api_response_query_daily_sales_report = {"date": "2023-09-29", "summary": "Total Sales Amount: 10000, Total Units Sold: 250"} # this needs to be a dictionary!!

# Append tool results
conversation.append({"role": "tool", "tool_call_id": "0", "content": api_response_query_daily_sales_report})

On 29th September 2023, the total sales amount was <co>10000</co: 0:[0]> and the total units sold were <co>250.</co: 0:[0]>

Model Card Contact

For errors or additional questions about details in this model card, contact [ [email protected] ].

Try it now:

You can try Command A+ in the playground . You can also use it in our dedicated Hugging Face Space .