Tight fit

Running DeepSeek R1 Distill Qwen 32B on NVIDIA GeForce RTX 4090

DeepSeek R1 Distill Qwen 32B fits at Q4_K_M, but headroom is tight (0 GB). Reduce context or use a smaller quant for safety.

By Fredoline Eruo·Last verified May 6, 2026

Memory available

Recommended quant

Q4_K_M
Highest quality that fits

Quick start with Ollama

1. Install
ollama pull deepseek-r1:32b
2. Run
ollama run deepseek-r1:32b

Default quant in Ollama is Q4_K_M. To use a different quant, append it: deepseek-r1:32b-q5_K_M.

Variants and what fits

QuantizationFile sizeVRAM requiredFits on NVIDIA GeForce RTX 4090?
Q4_K_M19.0 GB24 GB
Yes
Q8_034.0 GB40 GB
No

Real benchmarks

ToolQuantContexttok/sVRAM usedSource
vLLMAWQ-INT432,76832.5 tok/s22.4 GB
community

Frequently asked

Can NVIDIA GeForce RTX 4090 run DeepSeek R1 Distill Qwen 32B?

DeepSeek R1 Distill Qwen 32B fits at Q4_K_M, but headroom is tight (0 GB). Reduce context or use a smaller quant for safety.

What quantization should I use?

Q4_K_M is the highest-quality variant of DeepSeek R1 Distill Qwen 32B that fits in 24 GB VRAM. Lower-bit quants will be smaller but lose some quality.

How fast will it be?

Measured at 32.5 tok/s on this combination (community-sourced).

See also: DeepSeek R1 Distill Qwen 32B, NVIDIA GeForce RTX 4090, all benchmarks.

Reviewed by RunLocalAI Editorial. See our editorial policy.