Skip to content

DeepSeek R1 Distill Qwen 7B VRAM requirements

How much VRAM DeepSeek R1 Distill Qwen 7B needs
4-bit weights: 4.3 GBWith 8K context: 5.6 GBWith 32K context: 6.9 GB8-bit, 8K context: 8.8 GB
Parameters
7.6B
Longest context
128K
Released
Jan 2025
License
MIT
Strongest at
Not rated

How much VRAM DeepSeek R1 Distill Qwen 7B needs

Memory DeepSeek R1 Distill Qwen 7B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit14.2 GB15.5 GB16.8 GB22.0 GB
8-bit7.5 GB8.8 GB10.1 GB15.4 GB
6-bit5.8 GB7.1 GB8.4 GB13.7 GB
5-bit5.0 GB6.3 GB7.6 GB12.9 GB
4-bit4.3 GB5.6 GB6.9 GB12.1 GB
3-bit3.5 GB4.8 GB6.1 GB11.3 GB
2-bit2.7 GB3.9 GB5.3 GB10.5 GB
1.6-bit1.5 GB2.7 GB4.1 GB9.3 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run DeepSeek R1 Distill Qwen 7B

The best way DeepSeek R1 Distill Qwen 7B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40606-bit16K32 tok/s1.0 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4060 report
RTX 50606-bit16K52 tok/s0.8 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5060 report
RTX 30706-bit16K52 tok/s0.7 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 3070 report
RTX 3080 10 GB8-bit20K67 tok/s0.5 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit52K33 tok/s1.1 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 3060 12 GB report
Arc B5808-bit52K29 tok/s0.8 sPDF: DeepSeek R1 Distill Qwen 7B on Arc B580 report
RTX 40708-bit52K45 tok/s0.5 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4070 report
RTX 4070 Super8-bit52K45 tok/s0.4 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4070 Super report
RTX 50708-bit52K59 tok/s0.5 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit120K26 tok/s0.7 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit120K40 tok/s0.6 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit120K35 tok/s0.7 sPDF: DeepSeek R1 Distill Qwen 7B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit120K48 tok/s0.4 sPDF: DeepSeek R1 Distill Qwen 7B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit120K59 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4070 Ti Super report
RTX 4080 Super8-bit120K65 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4080 Super report
RTX 5070 Ti8-bit120K78 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5070 Ti report
RTX 50808-bit120K83 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5080 report
Radeon RX 7900 XTX8-bit128K69 tok/s0.6 sPDF: DeepSeek R1 Distill Qwen 7B on Radeon RX 7900 XTX report
RTX 30908-bit128K81 tok/s0.4 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 3090 report
RTX 40908-bit128K87 tok/s0.2 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 4090 report
Radeon AI PRO R97008-bit128K48 tok/s0.4 sPDF: DeepSeek R1 Distill Qwen 7B on Radeon AI PRO R9700 report
RTX 50908-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 5090 report
2× RTX 30908-bit128K81 tok/s0.2 sPDF: DeepSeek R1 Distill Qwen 7B on 2× RTX 3090 report
2× RTX 40908-bit128K87 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit128K113 tok/s0.2 sPDF: DeepSeek R1 Distill Qwen 7B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit128K83 tok/s0.2 sPDF: DeepSeek R1 Distill Qwen 7B on RTX 6000 Ada report
L40S8-bit128K75 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on L40S report
2× RTX 50908-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 2× RTX 5090 report
4× RTX 30908-bit128K81 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit32K13 tok/s9.2 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit128K13 tok/s9.2 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit128K13 tok/s9.2 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit128K28 tok/s4.6 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit128K28 tok/s4.6 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit128K28 tok/s4.6 sPDF: DeepSeek R1 Distill Qwen 7B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit128K49 tok/s2.3 sPDF: DeepSeek R1 Distill Qwen 7B on M4 Max 128 GB report
M5 Max 128 GB8-bit128K54 tok/s0.7 sPDF: DeepSeek R1 Distill Qwen 7B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit128K61 tok/s1.3 sPDF: DeepSeek R1 Distill Qwen 7B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit128K61 tok/s1.3 sPDF: DeepSeek R1 Distill Qwen 7B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit128K61 tok/s1.3 sPDF: DeepSeek R1 Distill Qwen 7B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit128K84 tok/s0.4 sPDF: DeepSeek R1 Distill Qwen 7B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit128K20 tok/s1.3 sPDF: DeepSeek R1 Distill Qwen 7B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit128K20 tok/s1.3 sPDF: DeepSeek R1 Distill Qwen 7B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit128K27 tok/s0.5 sPDF: DeepSeek R1 Distill Qwen 7B on RTX Spark report
DGX Spark8-bit128K25 tok/s0.5 sPDF: DeepSeek R1 Distill Qwen 7B on DGX Spark report
Dual DGX Spark8-bit128K25 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on Dual DGX Spark report
DGX Station GB3008-bit128K425 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit128K146 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit128K164 tok/s0.3 sPDF: DeepSeek R1 Distill Qwen 7B on A100 80 GB SXM report
H100 SXM8-bit128K247 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on H100 SXM report
4× H200 NVL8-bit128K325 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 4× H200 NVL report
8× H200 NVL8-bit128K325 tok/s0.1 sPDF: DeepSeek R1 Distill Qwen 7B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf bartowski/DeepSeek-R1-Distill-Qwen-7B-GGUF
vLLM
vllm serve deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
MLX (Mac)
mlx_lm.chat --model mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit

Questions

How much VRAM does DeepSeek R1 Distill Qwen 7B need?

DeepSeek R1 Distill Qwen 7B needs 4.3 GB for its weights at 4-bit, 5.6 GB with an 8K context and 6.9 GB with 32K. At 8-bit, 8.8 GB.

What is the minimum VRAM for DeepSeek R1 Distill Qwen 7B?

About 4.8 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060.

How big is DeepSeek R1 Distill Qwen 7B?

7.6B parameters: 14.2 GB of weights at 16-bit, 4.3 GB at 4-bit.

Can DeepSeek R1 Distill Qwen 7B run on an RTX 4090?

Yes. At 8-bit it holds 128K of context and writes about 87 tokens a second.

More models