Can I run DeepSeek R1 671B locally?
Not on a single consumer GPU. Even at Q4 with 8K context, DeepSeek R1 671B needs about 380 GB. That takes several GPUs, a workstation card, or a machine with a lot of fast system RAM, where it runs slowly.
Memory needed by compression and context length
| Compression | 4K context | 8K context | 32K context | 128K context |
|---|---|---|---|---|
Q4 The most common download. Small quality loss. | 380 GB | 380 GB | 382 GB | 388 GB |
Q8 Practically indistinguishable from FP16 at about half the size. | 665 GB | 666 GB | 667 GB | 674 GB |
Total memory: model weights, context (KV cache) and about 1 GB of runtime overhead. To run entirely on the GPU, VRAM must be about 5% larger than this.
GPUs that run DeepSeek R1 671B entirely on the GPU
None of the GPUs in our database has the 401 GB of VRAM it needs at Q4.
Model facts
- Parameters
- 671B (37B active, mixture-of-experts)
- Max context
- 128K tokens
- License
- MIT
- Released
- Jan 2025
Check it with your hardware
Change the compression, context length and system RAM in the calculator.
Open calculator