Can I run DeepSeek V2 Chat locally?
DeepSeek V2 Chat by deepseek-ai has 236B parameters (21B active per token, 160 experts, 6 used per token). At 4-bit with an 8K context it needs about 135 GB of GPU memory. It fits on 9 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 83.2 GB | 83.5 GB | 85.0 GB | 91.4 GB |
| 3-bit (Q3_K_M) | usable | 108 GB | 108 GB | 110 GB | 116 GB |
| 4-bit (Q4_K_M) | good, most popular | 135 GB | 135 GB | 137 GB | 143 GB |
| 5-bit (Q5_K_M) | very good | 157 GB | 158 GB | 159 GB | 165 GB |
| 6-bit (Q6_K) | near lossless | 181 GB | 181 GB | 183 GB | 189 GB |
| 8-bit (Q8_0) | effectively lossless | 234 GB | 234 GB | 236 GB | 242 GB |
| 16-bit (F16/BF16) | original weights | 440 GB | 440 GB | 442 GB | 448 GB |
Which GPUs and Macs run DeepSeek V2 Chat?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Questions
How much VRAM does DeepSeek V2 Chat need?
About 135 GB at 4-bit (Q4_K_M) with an 8K context: 134 GB of weights (estimated), 0.5 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 234 GB.
What is the smallest GPU that runs DeepSeek V2 Chat?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the Mac M3 Ultra 256GB (256 GB).
Can I run DeepSeek V2 Chat on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M3 Ultra 256GB or larger, using the default macOS GPU memory limit.
Model facts
| Hugging Face repo | deepseek-ai/DeepSeek-V2-Chat |
| Parameters | 235,741,434,880 (236B) · ~21B active (computed from config) |
| Layers / hidden size | 60 / 5120 |
| Attention | MLA; 128 heads, 128 KV heads, head dim 40 |
| KV cache per token (FP16) | 67.5 KiB |
| Max context | 163,840 tokens |
| License | other |
| Released on HF | Apr 28, 2024 |
| Downloads (30d) / likes | 26,006 / 486 |
Sources: Hugging Face model API · config.json (deepseek-ai/DeepSeek-V2-Chat) · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.