Model
Model explorer

Command A

OPEN
Cohere · Command A family · released Mar 13, 2025

111B enterprise flagship, 256K context, runs on just two GPUs (A100/H100); rivals GPT-4o and DeepSeek-V3 on agentic enterprise tasks.

ReasoningCodingVisionFunction callingTool useAgentic
1177.5
Elo · rank #182
Parameters
111B
Active params
111B (dense)
Context
256K tokens
Architecture
Dense transformer (111B) with hybrid sliding-window/full attention (3:1 ratio) and GQA
License
CC-BY-NC-4.0 (+ Acceptable Use Policy)
Languages
23+
API price (in/out)
$2.5 / $10
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
BFCLAgents63.8%#25
best: SmolLM3 3B (No Thinking) · 92.3%
BIRD-SQLCoding59.5%#1
best: this model · 59.5%
GPQA DiamondReasoning50.8%#185
best: GPT-5.6 · 94.6%
IFEvalReasoning90.9%#12
best: Gemma 4 26B A4B · 98.5%
LiveCodeBenchCoding35.1%#140
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MATH-500Math80.0%#69
best: GPT-5 · 99.4%
MBPP+Coding86.2%#4
best: Llama 3.1 405B · 88.6%
MMLU-ProKnowledge69.2%#92
best: Claude Fable 5 · 91.5%
MMLUKnowledge85.5%#53
best: OpenAI o3 · 92.9%
SWE-bench VerifiedCoding7.8%#118
best: Claude Fable 5 · 95.0%
tau-benchAgents51.7%#20
best: Claude Opus 4.1 · 82.4%
Run it locally
VRAM @ Q4
67 GB
VRAM @ FP16
222 GB
Fits on (Q4)
M3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Requires ~2x A100/H100 GPUs for practical serving; does not fit on a single RTX 4090 even quantized
Quantizations
GGUF Q4
Command A family
Elo progression across releases
Command AMar 20251177.5Command A+May 20261749.0
Resources
HFCohereLabs/c4ai-command-a-03-2025DOCSdocs.cohere.comAPIAvailable via Cohere API
API price $2.5/$10 · each benchmark row carries its own source badge (see methodology)