RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can DeepSeek R1 run on the L40S?

No. Even at 2-bit with 8K context, DeepSeek R1 needs about 228 GB, more than the 45.6 GB usable on the L40S.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •228 GB No228 GB No230 GB No236 GB No—
3-bit (Q3_K_M) •298 GB No298 GB No300 GB No307 GB No—
4-bit (Q4_K_M) •378 GB No378 GB No379 GB No386 GB No—
5-bit (Q5_K_M) •444 GB No444 GB No446 GB No452 GB No—
6-bit (Q6_K) •514 GB No514 GB No516 GB No522 GB No—
8-bit (Q8_0) •665 GB No665 GB No667 GB No674 GB No—
16-bit (F16/BF16) •1251 GB No1251 GB No1253 GB No1259 GB No—

• real GGUF file size from unsloth/DeepSeek-R1-GGUF. Speed is a bandwidth ceiling (864 GB/s), not a benchmark. Methodology.

Hardware that does fit it

Mac M3 Ultra 512GB

More

DeepSeek R1 on every GPU and Mac · Every model that runs on the L40S · Other models that fit 48 GB VRAM · Try other settings in the calculator

Data: Hugging Face Hub (deepseek-ai/DeepSeek-R1), NVIDIA specs. Data as of 2026-11-17.