Can I run DeepSeek V3.2 Exp locally?
DeepSeek V3.2 Exp by deepseek-ai has 685B parameters (41B active per token, 256 experts, 8 used per token). At 4-bit with an 8K context it needs about 391 GB of GPU memory. It fits on 1 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 | 240 GB | 241 GB | 243 GB | 251 GB |
| 3-bit (Q3_K_M) | usable | 313 GB | 313 GB | 315 GB | 323 GB |
| 4-bit (Q4_K_M) | good, most popular | 391 GB | 391 GB | 393 GB | 401 GB |
| 5-bit (Q5_K_M) | very good | 456 GB | 456 GB | 458 GB | 466 GB |
| 6-bit (Q6_K) | near lossless | 524 GB | 525 GB | 527 GB | 535 GB |
| 8-bit (Q8_0) | effectively lossless | 679 GB | 680 GB | 681 GB | 689 GB |
| 16-bit (F16/BF16) | original weights | 1278 GB | 1278 GB | 1280 GB | 1288 GB |
Which GPUs and Macs run DeepSeek V3.2 Exp?
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).
Prefer an API?
DeepSeek: DeepSeek V3.2 Exp is on OpenRouter at $0.27 per 1M input tokens and $0.41 per 1M output tokens. Price details and history →
Questions
How much VRAM does DeepSeek V3.2 Exp need?
About 391 GB at 4-bit (Q4_K_M) with an 8K context: 390 GB of weights (estimated), 0.7 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 680 GB.
What is the smallest GPU that runs DeepSeek V3.2 Exp?
No single consumer GPU in our table fits it at Q4_K_M with 8K context.
Can I run DeepSeek V3.2 Exp on a Mac?
Not at Q4_K_M with 8K context on any Mac in our table, even a 512 GB Mac Studio, using the default GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
DeepSeek V3.2
685B · 379 GB at 4-bitDeepSeek V3.1
685B · 379 GB at 4-bitDeepSeek V3 0324
685B · 378 GB at 4-bitGLM 5.3
753B · 430 GB at 4-bitGLM 5.2
753B · 435 GB at 4-bitGLM 5
754B · 426 GB at 4-bitModel facts
| Hugging Face repo | deepseek-ai/DeepSeek-V3.2-Exp |
| Parameters | 685,396,921,376 (685B) · ~41B active (computed from config) |
| Layers / hidden size | 61 / 7168 |
| Attention | MLA; 128 heads, 128 KV heads, head dim 56 |
| KV cache per token (FP16) | 83.9 KiB |
| Max context | 163,840 tokens |
| License | mit |
| Released on HF | Sep 29, 2025 |
| Downloads (30d) / likes | 53,092 / 1,025 |
Sources: Hugging Face model API · config.json (deepseek-ai/DeepSeek-V3.2-Exp) · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.