Mac Studio M3 Ultra
Apple Silicon
Compatible with limitsRuns DeepSeek R1 (671B) Q4_K_M entirely in memory
Model · DeepSeek
DeepSeek R1 (671B): VRAM and hardware requirements
At Q4_K_M with an 8K context, DeepSeek R1 (671B) needs about 403 GB: a GPU with – or more runs it entirely in VRAM.
- Parameters671B
- Active37B
- Max context160K
- LicenseMIT
Specifications
- Organization
- DeepSeek
- Architecture
- deepseek_v3
- Released
- January 2025
- Layers
- 61
- Hidden size
- 7168
- Attention heads
- 128
- KV heads
- MLA (rank 512)
- Head dimension
- 192
- Vocabulary
- 129,280
- KV cache per 1K tokens (f16)
- 0.067 GB
- Repository
- deepseek-ai/DeepSeek-R1-0528
Official specData updated 24 days agoSource: Hugging Face config.json + model card
Quantizations and memory
| Quantization | File size | Memory needed | Minimum GPU memory | Quality |
|---|---|---|---|---|
| Q8_0 | 664 GB | ~699 GB | – | Near-lossless |
| Q6_K | 515 GB | ~540 GB | – | Near-lossless |
| Q5_K_M | 445 GB | ~469 GB | – | Very good |
| Q4_K_MMost used | 381 GB | ~403 GB | – | Good — the usual choice |
| IQ4_XS | 335 GB | ~367 GB | – | Good — the usual choice |
| Q3_K_M | 287 GB | ~329 GB | – | Noticeable loss |
| Q2_K | 222 GB | ~261 GB | – | Significant loss |
File sizes: not available until imported from the model repository. Memory = weights (estimated from bits per weight) + KV cache + runtime overhead + safety margin. (Q8_0 ≈ 8.5 bpw, Q6_K ≈ 6.57 bpw, Q5_K_M ≈ 5.7 bpw, Q4_K_M ≈ 4.9 bpw, IQ4_XS ≈ 4.46 bpw, Q3_K_M ≈ 4 bpw, Q2_K ≈ 3.17 bpw)
Q4, Q5 or Q8: how to choose a quantization
GPUs that run DeepSeek R1 (671B)
| RTX 509032 GB · 64 GB RAM~402 GB · No data | Incompatible |
|---|---|
| RTX 409024 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 309024 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RX 7900 XTX24 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 5060 Ti 16 GB16 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 5070 Ti16 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 508016 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 4060 Ti 16 GB16 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 4070 Ti SUPER16 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RX 9070 XT16 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 3060 12 GB12 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 407012 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 507012 GB · 64 GB RAM~402 GB · No data | Incompatible |
| RTX 40608 GB · 64 GB RAM~402 GB · No data | Incompatible |
Macs that run DeepSeek R1 (671B)
| M3 Ultra (80-core GPU) · 512 GBMac Studio~401 GB · No data | Compatible with limits |
|---|---|
| M4 Max (40-core GPU) · 128 GBMacBook Pro 14, MacBook Pro 16~401 GB · No data | Incompatible |
| M3 Max (40-core GPU) · 128 GBMacBook Pro 14, MacBook Pro 16~401 GB · No data | Incompatible |
| M4 Pro (20-core GPU) · 64 GBMac mini, MacBook Pro 14~401 GB · No data | Incompatible |
| M4 Max (32-core GPU) · 36 GBMacBook Pro 14, MacBook Pro 16~401 GB · No data | Incompatible |
| M4 · 32 GBMacBook Air, MacBook Pro 14~401 GB · No data | Incompatible |
| M5 · 32 GBMacBook Pro 14~401 GB · No data | Incompatible |
| M3 · 24 GBMacBook Air, MacBook Pro 14~401 GB · No data | Incompatible |
EstimateData updated 24 days agoSource: Run LLM Locally engine
Limits in plain words
DeepSeek R1 (671B) is a mixture-of-experts model: all 671B parameters must be in memory, but only 37B are read per token, so it generates much faster than a dense model of the same size.
The KV cache grows with context: at 32K it adds about 2.1 GB on top of the weights.
With under 8 GB of VRAM, use a smaller quantization or a smaller model of the same family.
Where to buy
Prices and links follow the market you choose.
No current offer imported for this market: the link opens a marketplace search.
Frequently asked questions
How much VRAM does DeepSeek R1 (671B) need?
About 403 GB at Q4_K_M with an 8K context, including the KV cache and runtime overhead. At Q8_0 it needs about 699 GB. A longer context adds more: about 2.1 GB of KV cache at 32K.
Can DeepSeek R1 (671B) run on a Mac?
Yes on a Mac whose GPU can use at least 401 GB, e.g. M3 Ultra (80-core GPU) 512 GB. macOS lets the GPU use about two thirds to three quarters of unified memory by default.