whatmodel.sh / Command A (03-2025)

Command A (03-2025)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Command A (03-2025) is an open-weights language model from Cohere, released March 2025. Its published size is 111B parameters and a 256k-token context window. Licence: cc-by-nc-4.0.

57.2±1.0#232 of 719 LLMsfrom 22 benchmark results across 6 sources

$ how it compares

The LLMs ranked either side of Command A (03-2025), out of 828 with a score, then more from Cohere.

#modelscorewhereRAMreleased
275DeepSeek R1 0528 Qwen3 8B DeepSeek57±1api $0.068/M +local~6.3GB2025-05
276Granite 4.2 3B IBM57±1api $0.052/M +local~3.7GB2026-08
277Command A (03-2025) Cohere57±1api $4.38/M +local~37GB2025-03
278GPT-4o (2024-11-20) OpenAI57±1api $4.38/M2024-11
279Gemini 2.5 Flash-Lite Google57±2api $0.17/M2025-07
more from Cohere
76North Small Translate 1.0 Cohere71±6api price unknown2026-08
175Command A+ (05-2026) Cohere64±1api price unknown +local~71GB2026-05
324Command A Reasoning (08-2025) Cohere55±2local~34GB

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #185 of 300 · 27% of its score · 6 tests

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 0528 here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified lab26.8% ✎▃···▃·▃·▃█·▃ beats 3/7hard · avg 19%▇▇▇▇ 11%sharp · at its level
BFCL V4 (December 2025 archive)46.5% ✎▁▁▁▂██▅▅▆▂▂▂ beats 52/77hard · avg 32%▇▇ 4.4%at its level
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)4.6% ✎█·▃▃·█▅▅▅▅▃▅ beats 3/18very hard · avg 5.8%▇▇ 4.3%sharp
τ-bench Airline lab45.3% ✎▁▁▁·▅▄▁█▇▁▁▇ beats 14/28hard · avg 23%▇ 2.7%at its level
τ-bench Retail lab60.0% ✎▁····▃·▂▇█▂▆ beats 9/28hard · avg 29%▇ 2.0%at its level · noisy
BFCL v364.4% ✎▁▁▁▁▁▃▅█▅▄▁▁ beats 95/114medium · avg 52%▇ 1.9%blunt · at its level
coding▁▁▁▃▆██▇▆▄▃▁ #220 of 458 · 38% of its score · 5 tests

Beats Qwen3.5 4B, Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B) here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 3B Reasoning 2512, DeepSeek-Coder-V2 Instruct (0724) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Aider polyglot (225 exercises, two attempts) lab14.7% ✎█▄··█·█····▄ beats 6/7very hard · avg 6.7%▇▇▇▇▇ 12%sharp · noisy
SWE-bench Verified lab26.8% ✎▃···▃·▃·▃█·▃ beats 3/7hard · avg 19%▇▇▇▇ 11%sharp · at its level
LiveCodeBench v5 (Oct 2024–Feb 2025) lab26.9% ✎▄▄·▄·····█▄█ beats 4/7hard · avg 19%▇▇▇ 7.1%at its level
MBPP+ lab86.2% ✎██·██···████ beats 4/7easy · avg 79%▇▇ 3.8%typical
BIRD-SQL (dev) lab59.5% ✎▄··▄·▄·█··▄█ beats 7/7medium · avg 49%▇ 3.3%at its level
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #154 of 393 * · 8.9% of its score · 2 tests

Beats MiniMax-M2, MiniCPM5 2B, Granite 4.2 30B here, though they rank higher overall. Trails LFM2.5 1.2B Thinking, LFM2.5 8B-A1B, LFM2.5 2.6B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1353.7 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 199/379avg 1268▇▇▇▇ 8.9%blunt
IFEval strict, mean of prompt and instruction level lab90.9% ✎█··········█ beats 1/1—not counted yet: no other model in the same run to compare it with
math▁▂▃▅▆▅▄█▄▁▁▁ #243 of 398 * · 9.0% of its score · 2 tests

Trails DeepSeek R1 Distill Qwen 1.5B, Tulu 2 DPO 70B, OpenThinker 7B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
AIME 2024 lab23.3% ✎▁▁▁▂▁▃▁▅███▃ beats 13/85hard · avg 16%▇▇▇▇ 9.0%sharp · at its level · small run
MATH lab80.0% ✎▃▅▂▆▅▃▅▃█▃▃▃ beats 29/29medium · avg 64%—not counted yet: no other model in the same run to compare it with
web search▂▁▂▆▄█▃▃▁▁·▁ #44 of 80 * · 4.3% of its score · 1 test

Beats Gemini 3.1 Pro (Preview), Qwen3.7 Max, GPT-5.2 here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)4.6% ✎█·▃▃·█▅▅▅▅▃▅ beats 3/18very hard · avg 5.8%▇▇ 4.3%sharp
other tests11 not in a task · 29% of its score
testresultvs the others in its rundifficultycountswhy
RepoQA (32K, threshold 0.8) lab92.6% ✎▄▄▄··▄▄··▄·█ beats 7/7easy · avg 81%▇▇ 5.0%typical
GPQA Diamond (simple-evals) lab50.8% ✎▁·····▄█▂▂▃▃ beats 11/22hard · avg 36%▇▇ 4.9%sharp · at its level · small run
BigCodeBench (Instruct) lab45.4% ✎▄··▄▄···▄▄█▄ beats 5/7medium · avg 36%▇▇ 4.5%at its level
MMLU-Pro lab69.6% ✎▁▁·▁▁▁▃▂▄▅█▁ beats 86/188medium · avg 56%▇ 3.6%at its level · small run
Open Persian LLM Leaderboard v2 average (21 tasks)70.0% ✎▂▄▂▃█▄▃▆▃▅▂▂ beats 72/81medium · avg 58%▇ 3.5%blunt · at its level
LongBench v2 lab43.4% ✎▂·▁·▂▂▁▃▄▁█▄ beats 11/28very hard · avg 34%▇ 3.4%small run
RULER 128k lab90.0% ✎▃··▅▄▇▅█▇·▃▃ beats 30/32easy · avg 70%▇ 1.9%small run
MMLU (simple-evals, zero-shot CoT) lab85.5% ✎▁···▂▁▅█▂▂▃▁ beats 9/22easy · avg 78%▇ 1.5%small run
3 more tests
testresultvs the others in its rundifficultycountswhy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)95.5% ✎▁▁▂▃▄▄█▆▃▂▂▁ beats 74/140very easy · avg 94%▇ 0.4%blunt · maxed out · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)90.7% ✎▁▂▁▂▅█▂▃▄▃▁▁ beats 56/98very easy · avg 90%▇ 0.0%blunt
InFoBench lab94.9% ✎█··········█ beats 1/1—not counted yet: no other model in the same run to compare it with

$ where to run it

Through an API

api · Microsoft Azure · $4.38/M · $2.5 in / $10 out per 1M tokens · 256k context · azure.microsoft.com ↗ · price sourceUSD per 1M tokens, Azure AI Foundry Models pay-as-you-go, 'Cohere Command A Global' (Azure model id cohere-command-a, version 1, GA with no retirement date on Azure's schedule): input $2.50, output $10.00, region US East. Cohere's own pricing page lists no Command A per-token rate. Currency: USD. Microsoft Azure official pricing page (price cells carry the amounts in the page's own data), observed 2026-09-25.

llama.cpp · recommended

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run Command A (03-2025) · needs ~33.8 GB · from huggingface.co/unsloth
    chat in the terminal
    llama cli -hf unsloth/c4ai-command-a-03-2025-GGUF:Q2_K
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf unsloth/c4ai-command-a-03-2025-GGUF:Q2_K

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Command A (03-2025) · needs ~62.5 GB, Mac only · from huggingface.co/mlx-community

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/mlx-community/c4ai-command-a-03-2025-4bit -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Command A (03-2025) · needs ~67.1 GB · from ollama.com
    chat in the terminal
    ollama run command-a:111b

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Command A (03-2025) · needs ~62.5 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/c4ai-command-a-03-2025-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/c4ai-command-a-03-2025-4bit

Weights and documentation: huggingface.co/CohereLabs ↗

$ how the score was worked out · Cohere — Command A technical… 76%, LMArena leaderboard dataset 9%, Berkeley Function Calling Lea… 6% and 3 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Cohere — Command A technical report · lab-reported15▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 76%CC BY 4.0
LMArena leaderboard dataset1▇▇▇ 9%CC BY 4.0
Berkeley Function Calling Leaderboard result archive2▇▇ 6%Apache-2.0
Finance Agent Benchmark (Vals AI, arXiv:2508.00828v1)1▇ 4%CC-BY-4.0
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)1▇ 4%Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)2▇ 0%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 57.2

not counted (3)

Cohere — Command A technical reportIFEval strict, mean of prompt and instruction level [correct]90.9%fewer than three models have this test yet, or this source ran it for this model alone
Cohere — Command A technical reportInFoBench [correct]94.9%fewer than three models have this test yet, or this source ran it for this model alone
Cohere — Command A technical reportMATH [correct]80.0%fewer than three models have this test yet, or this source ran it for this model alone

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json