Model
Model explorer

Command R

OPEN
Cohere · Command R family · released Mar 11, 2024

35B RAG/tool-use model with 128K context; first open-weights (CC-BY-NC) Command release, later superseded by the 08-2024 refresh.

ReasoningCodingVisionFunction callingTool useAgentic
385.9
Elo · rank #343
Parameters
35B
Active params
35B (dense)
Context
128K tokens
Architecture
Dense transformer (35B), optimized for RAG and multi-step tool use
License
CC-BY-NC-4.0 (+ Acceptable Use Policy)
Languages
10+
API price (in/out)
$0.5 / $1.5
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning65.5%#62
best: Llama 3.1 405B · 96.9%
BIG-Bench HardReasoning53.5%#90
best: ERNIE 4.5 300B-A47B · 94.3%
GPQA DiamondReasoning30.3%#270
best: GPT-6 Astra · 96.0%
GSM8KMath56.6%#122
best: Llama 3.1 405B · 96.8%
HellaSwagReasoning87.0%#20
best: Claude 3 Opus · 95.4%
IFEvalReasoning62.7%#119
best: Gemma 4 26B A4B · 98.5%
MMLU (EU-21 languages)Knowledge56.1%#7
best: Llama 3.1 70B · 77.1%
MMLU-ProKnowledge33.7%#161
best: Claude Fable 5 · 91.5%
MMLUKnowledge68.2%#158
best: OpenAI o3 · 92.9%
TruthfulQAKnowledge52.3%#48
best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%
WinoGrandeReasoning81.5%#26
best: PaLM 2 · 90.9%
Run it locally
VRAM @ Q4
21.6 GB
VRAM @ FP16
70 GB
Fits on (Q4)
RTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF Q4
Fine-tune it
Research only
QLoRA24.1 GB1× RTX 5090 32GB
LoRA74.8 GB1× A100 80GB
Full fine-tune562.0 GB4× H200 141GB
QLoRA SFT on ~10k samples ≈ $17.07 (1× RTX 5090 32GB)
Plan a fine-tune →
Command R family
Elo progression across releases
Command RMar 2024385.9Command R7BDec 2024491.3
Resources
HFCohereLabs/c4ai-command-r-v01DOCSdocs.cohere.comAPIAvailable via Cohere API
API price $0.5/$1.5 · each benchmark row carries its own source badge (see methodology)