DeepSeek V3.2
672B total, 38.5B active · 160k context · MIT License · released 2025-12-01
chatreasoningagentic
In short
DeepSeek V3.2 is an open model from DeepSeek with 672B parameters (38.5B active per token). At Q4_K_M, DeepSeek V3.2 needs about 379 GB of memory at 8k context and 388 GB at 128k, counting the weights, the KV cache and runtime buffers. Nothing we price in the UK runs DeepSeek V3.2 fully in memory at 128k context at a usable speed; see the table for what comes closest.
Updated . Speeds are estimates; how they are worked out.
What to run it on in the UK
Fully in memory at 128k context, priced at retailers today.
Nothing we price holds this model fully in memory.
All hardware that runs it
Everything we price, best value first. Two cards split the model by layers.
Nothing we price can run it at this context.
Memory needed
Weights, KV cache and runtime buffers at each context length (KV cache at fp16).
| Quant | File | @8k | @32k | @64k | @128k |
|---|---|---|---|---|---|
| Q4_K_Mgguf | 377.6 GB | 378.8 GB | 380.6 GB | 382.9 GB | 387.6 GB |
| mlx-5bitmlx | 430.3 GB | 431.6 GB | 433.4 GB | 435.7 GB | 440.4 GB |
| Q5_K_Mgguf | 443.5 GB | 444.7 GB | 446.5 GB | 448.8 GB | 453.5 GB |
| Q6_Kgguf | 513.5 GB | 514.7 GB | 516.5 GB | 518.8 GB | 523.5 GB |
| Q8_0gguf | 664.3 GB | 665.6 GB | 667.3 GB | 669.7 GB | 674.3 GB |
Run it
Copy a command for the runtime you use.
- Q4_K_M
- Ollama
ollama run hf.co/unsloth/DeepSeek-V3.2-GGUF:Q4_K_M - LM Studio
lms get unsloth/DeepSeek-V3.2-GGUF@q4_k_m - llama.cpp
llama-server -hf unsloth/DeepSeek-V3.2-GGUF:Q4_K_M
- Ollama
- mlx-5bit
- LM Studio
lms get mlx-community/DeepSeek-V3.2-mlx-5bit - MLX
mlx_lm.server --model mlx-community/DeepSeek-V3.2-mlx-5bit
- LM Studio
- Q5_K_M
- Ollama
ollama run hf.co/unsloth/DeepSeek-V3.2-GGUF:Q5_K_M - LM Studio
lms get unsloth/DeepSeek-V3.2-GGUF@q5_k_m - llama.cpp
llama-server -hf unsloth/DeepSeek-V3.2-GGUF:Q5_K_M
- Ollama
- Q6_K
- Ollama
ollama run hf.co/unsloth/DeepSeek-V3.2-GGUF:Q6_K - LM Studio
lms get unsloth/DeepSeek-V3.2-GGUF@q6_k - llama.cpp
llama-server -hf unsloth/DeepSeek-V3.2-GGUF:Q6_K
- Ollama
- Q8_0
- Ollama
ollama run hf.co/unsloth/DeepSeek-V3.2-GGUF:Q8_0 - LM Studio
lms get unsloth/DeepSeek-V3.2-GGUF@q8_0 - llama.cpp
llama-server -hf unsloth/DeepSeek-V3.2-GGUF:Q8_0
- Ollama
Notes
No Jinja chat template is included; use the provided Python encoding scripts (encoding_dsv32.py). Recommended sampling: temperature 1.0, top_p 0.95. Sparse or compressed attention: KV read per token is estimated as full attention. Excludes 13.4B parameters of multi-token-prediction layers, which local runtimes do not load.
Questions
- How much VRAM does DeepSeek V3.2 need?
- At Q4_K_M, DeepSeek V3.2 needs about 379 GB of memory at 8k context and 388 GB at 128k, counting the weights, the KV cache and runtime buffers.
- Can I run DeepSeek V3.2 on a 16 GB GPU?
- No: it does not fit in 16 GB at 128k context, even with system RAM.
- Can I run DeepSeek V3.2 on a 24 GB GPU?
- No: it does not fit in 24 GB at 128k context, even with system RAM.
- How do I run DeepSeek V3.2 locally?
- With Ollama: ollama run hf.co/unsloth/DeepSeek-V3.2-GGUF:Q4_K_M (the Q4_K_M build). LM Studio and llama.cpp commands are on this page.
- What licence is DeepSeek V3.2 released under?
- MIT License, which allows commercial use.