FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 34.4 GB
31.4 + 3.0 KV
low IQ2_M 2.93 41.6 GB
38.6 + 3.0 KV
low Q2_K 3.16 44.6 GB
41.6 + 3.0 KV
low IQ3_XXS 3.25 45.7 GB
42.7 + 3.0 KV
low IQ3_XS 3.5 49.0 GB
46.0 + 3.0 KV
low Q3_K_S 3.64 50.8 GB
47.8 + 3.0 KV
low IQ3_M 3.76 52.4 GB
49.4 + 3.0 KV
low Q3_K_M 4 55.5 GB
52.5 + 3.0 KV
low Q3_K_L 4.3 59.4 GB
56.4 + 3.0 KV
moderate IQ4_XS 4.46 61.5 GB
58.5 + 3.0 KV
moderate Q4_K_S 4.67 64.2 GB
61.2 + 3.0 KV
moderate Q4_K_M 4.89 67.1 GB
64.1 + 3.0 KV
good Q5_K_S 5.57 75.9 GB
72.9 + 3.0 KV
good Q5_K_M 5.7 77.6 GB
74.6 + 3.0 KV
good Q6_K 6.56 88.8 GB
85.8 + 3.0 KV
excellent Q8_0 8.5 114.0 GB
111.0 + 3.0 KV
lossless FP16 16 211.5 GB
208.5 + 3.0 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise → ▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN COMMAND-R+ 104B NOW
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Community Ratings Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model IQ2_XXS — 31.4 GB VRAM IQ2_M — 38.6 GB VRAM Q2_K — 41.6 GB VRAM IQ3_XXS — 42.7 GB VRAM IQ3_XS — 46.0 GB VRAM Q3_K_S — 47.8 GB VRAM IQ3_M — 49.4 GB VRAM Q3_K_M — 52.5 GB VRAM Q3_K_L — 56.4 GB VRAM IQ4_XS — 58.5 GB VRAM Q4_K_S — 61.2 GB VRAM Q4_K_M — 64.1 GB VRAM Q5_K_S — 72.9 GB VRAM Q5_K_M — 74.6 GB VRAM Q6_K — 85.8 GB VRAM Q8_0 — 111.0 GB VRAM FP16 — 208.5 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run command-r-plus:104b-08-2024-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ SETUP GUIDE >_
Auto-setup with fitmyllm CLI Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for Command-R+ 104B
Build Hardware for Command-R+ 104B Command R+ — Cohere's flagship open model. Excellent for RAG and tool use.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Command-R+ 104B — 104B Dense. ▸ SPECIFICATIONS
PARAMETERS 104B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 128K tokens
CAPABILITIES chat, reasoning
RELEASE DATE 2024-04-04
PROVIDER Cohere
FAMILY command ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 31.4 GB 65% IQ2_M 2.93 38.6 GB 75% Q2_K 3.16 41.6 GB 78% IQ3_XXS 3.25 42.7 GB 82% IQ3_XS 3.5 46.0 GB 84% Q3_K_S 3.64 47.8 GB 85% IQ3_M 3.76 49.4 GB 86% Q3_K_M 4 52.5 GB 88% Q3_K_L 4.3 56.4 GB 90% IQ4_XS 4.46 58.5 GB 92% Q4_K_S 4.67 61.2 GB 93% Q4_K_M 4.89 64.1 GB 94% Q5_K_S 5.57 72.9 GB 96% Q5_K_M 5.7 74.6 GB 96% Q6_K 6.56 85.8 GB 97% Q8_0 8.5 111.0 GB 100% FP16 16 208.5 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 81.0
MMLU-PRO 56.0
MATH 47.2
IFEval 85.4
BBH 74.0
GPQA 13.4
MUSR 19.8
MBPP 63.5
BigCodeBench 33.8
Arena Elo 1230.0
§ 02 RUN COMMAND
Run Command-R+ 104B locally with Ollama — needs 64.1 GB VRAM at Q4_K_M:
$ ollama run command-r-plus:104b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 