Cohere on NVIDIA Ada

Can I run Command R 35B on an RTX 4060 Ti 16GB?

Not at Q4_K_M — it needs 31.0 GB against 14.2 GB available. You would need 3 of these cards.

Won't fit

31.0 GB of 14.2 GB · 217%
032 GB
Weights 20.1 GB
KV cache 10.0 GB
Runtime overhead 0.9 GB
Over the limit 16.7 GB

Short by 16.7 GB. You can run it with 6 of 40 layers on the RTX 4060 Ti 16GB and the rest in system RAM, at roughly 1.73 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation1.73tok/s
Prompt processing264tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Command R 35B on a RTX 4060 Ti 16GB

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 65.2 GB 76.0 GB Won't fit 0.57
INT8 / W8A8 8.50 33.7 GB 44.6 GB Won't fit 1.05
Q8_0 (GGUF) 8.50 33.7 GB 44.6 GB Won't fit 1.05
FP8 (E4M3) 8.00 32.6 GB 43.5 GB Won't fit 1.10
Q6_K 6.56 26.7 GB 37.6 GB Won't fit 1.34
Q5_K_M 5.67 23.1 GB 34.0 GB Won't fit 1.51
AWQ 4-bit 4.25 23.0 GB 33.9 GB Won't fit 1.51
GPTQ 4-bit 4.25 23.0 GB 33.9 GB Won't fit 1.51
MXFP4 4.25 23.0 GB 33.9 GB Won't fit 1.51
Q5_K_S 5.52 22.5 GB 33.4 GB Won't fit 1.54
Q4_K_M 4.85 20.1 GB 31.0 GB Won't fit 1.73
Q4_K_S 4.58 19.1 GB 30.0 GB Won't fit 1.84
Q4_0 4.55 19.0 GB 29.9 GB Won't fit 1.85
IQ4_XS 4.25 17.9 GB 28.8 GB Won't fit 1.93
Q3_K_M 3.91 16.7 GB 27.6 GB Won't fit 2.09
IQ3_M 3.70 15.9 GB 26.8 GB Won't fit 2.16
IQ3_XXS 3.06 13.7 GB 24.5 GB Won't fit 2.49
Q2_K 2.63 12.1 GB 23.0 GB Won't fit 896 2.84
IQ2_XXS 2.06 10.1 GB 21.0 GB Won't fit 3K 3.40
IQ1_M 1.75 9.0 GB 19.8 GB Won't fit 4K 3.77

Also worth checking