~/can-i-run/command-r-plus
Cohere

Can I run Command R+?

Short answer: yes, on a M3 Ultra 512 (512GB) at FP16/BF16. Long answer below.

command-r-plusLM StudiovLLMMLXoMLX

The math, in one paragraph.

$ ./vrambudget --explain command-r-plus

Command R+ has 104B parameters. At FP16 that's 208 GB of raw weights. Quantization shrinks that, but you also need budget for the KV cache (definition), framework overhead, and safety headroom. The rule of thumb: real usable budget on a card is roughly its nameplate VRAM minus 25%. That's how the table below was computed.

What hardware actually fits.

$ grep "fits" gpus.json
FP16/BF16
208GB
1 GPU fits
M3 Ultra 512512GB
Q8_0
111GB
6 GPUs fit
H200141GB2× H100 NVL188GBM2 Ultra 192192GBB200192GB+ 2 more
Q5_K_M
72GB
12 GPUs fit
M3 Max 9696GBRTX Pro 600096GBM4 Max 128128GBM5 Max 128128GB+ 8 more
Q4_K_M
59GB
13 GPUs fit
H100 80GB80GBM3 Max 9696GBRTX Pro 600096GBM4 Max 128128GB+ 9 more
Q3_K_M
45GB
17 GPUs fit
M2 Max 6464GBM3 Max 6464GBM4 Pro 6464GBM5 Pro 6464GB+ 13 more

Pick your path.

$ ls strategies/
Tightest budget

Smallest GPU that fits Command R+ at any quant: M3 Ultra 512 at FP16/BF16.

Reference quality (FP16)

Lossless inference needs 208 GB. Pick from 1 cards.

Best quality on a 24GB card

None of the showcase quants fit on a 24GB card. Step up.

Tune the math yourself

Open the calculator pre-tuned for Command R+: ↗ /calc?model=command-r-plus

See the full model page.

$ ./open
$ openCommand R+104BCohere · ctx 128K$ rankBest GPU for Command R+see the ranking →ranked by fit quality

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.