Skip to content

Command R+ 104B

by Cohere · command-r family

104B

parameters

text-generation code-generation reasoning multilingual tool-use summarization

Command R+ 104B is Cohere's largest open-weight model, designed for enterprise retrieval-augmented generation, tool use, and multilingual workflows. It builds on Command R's strengths with significantly improved reasoning, accuracy, and instruction following at 104B parameters. The model supports 128K context and excels at grounded generation with source citations, multi-step tool use, and complex document analysis across 10+ languages. Its size requires substantial hardware but delivers near-frontier performance for RAG and agentic use cases.

Quick Start with Ollama

ollama run q4_K_M
Resources Ollama Hugging Face Official Page
Creator Cohere
Parameters 104B
Architecture transformer-decoder
Context 128K tokens
Released Apr 4, 2024
License CC-BY-NC-4.0
Ollama command-r-plus

Quantization Options

Format File Size VRAM Required Quality Ollama Tag
Q4_K_M rec 52 GB 57 GB q4_K_M
Q8_0 104 GB 110 GB q8_0

Compatible Hardware

Q4_K_M requires 57 GB VRAM

Compatible Hardware

HardwareVRAMTypeFitEst. Speed
Mac Studio M4 Ultra 512GB512 GBmacRuns~14 tok/s
Mac Pro M2 Ultra 192GB192 GBmacRuns~14 tok/s
Mac Studio M4 Ultra 192GB192 GBmacRuns~14 tok/s
Mac Studio M4 Max 128GB128 GBmacRuns~10 tok/s
MacBook Pro M4 Max 128GB128 GBmacRuns~10 tok/s
MacBook Pro M5 Max 128GB128 GBmacRuns~10 tok/s
NVIDIA RTX PRO 6000 Blackwell96 GBgpuRuns~34 tok/s
MacBook Pro M3 Max 96GB96 GBmacRuns~7 tok/s
Mac mini M4 Pro 64GB64 GBmacRuns (tight)~5 tok/s
Mac Studio M4 Max 64GB64 GBmacRuns (tight)~10 tok/s
MacBook Pro M4 Max 64GB64 GBmacRuns (tight)~10 tok/s
MacBook Pro M5 Max 64GB64 GBmacRuns (tight)~10 tok/s
NVIDIA RTX 6000 Ada Generation48 GBgpuCPU Offload~5 tok/s
NVIDIA RTX A600048 GBgpuCPU Offload~4 tok/s
NVIDIA RTX PRO 5000 Blackwell48 GBgpuCPU Offload~5 tok/s
Mac mini M4 Pro 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M3 Max 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M4 Pro 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M4 Max 48GB48 GBmacCPU Offload~3 tok/s
MacBook Pro M5 Max 48GB48 GBmacCPU Offload~2 tok/s
MacBook Pro M5 Pro 48GB48 GBmacCPU Offload~2 tok/s
86 hardware device(s) cannot run this model at Q4_K_M.

Benchmark Scores

75.7
mmlu