How much memory does DeepSeek V4 Pro need?
About 978 GB at Q4_K_M with an 8K context. That is more than any single Mac.
Try other settings in the calculatorMemory by quantization and context
Total memory: weights, KV cache in F16 and about 1 GB of buffers. Under each total, the smallest Mac that can run it with the default GPU memory limit.
| Quantization | 8K context | 32K context | 128K context |
|---|---|---|---|
| Q3_K_M | 782 GBtoo big for a Mac | 783 GBtoo big for a Mac | 783 GBtoo big for a Mac |
| Q4_K_M | 978 GBtoo big for a Mac | 979 GBtoo big for a Mac | 979 GBtoo big for a Mac |
| Q5_K_M | 1,140 GBtoo big for a Mac | 1,140 GBtoo big for a Mac | 1,141 GBtoo big for a Mac |
| Q6_K | 1,312 GBtoo big for a Mac | 1,312 GBtoo big for a Mac | 1,313 GBtoo big for a Mac |
| Q8_0 | 1,700 GBtoo big for a Mac | 1,700 GBtoo big for a Mac | 1,701 GBtoo big for a Mac |
| F16 | 3,199 GBtoo big for a Mac | 3,199 GBtoo big for a Mac | 3,200 GBtoo big for a Mac |
About the model
- Parameters
- 1,599B
- Type
- Mixture of experts, 384 experts
- Attention
- compressed sparse attention
- KV cache, F16
- 0.1 GB at 8K, 1.0 GB at 128K
- On Hugging Face
- deepseek-ai/DeepSeek-V4-Pro
Most layers store their cache only every few tokens or keep a short window, so the KV cache is a fraction of what a standard model would need. As a mixture-of-experts model it loads every expert into memory, so memory follows the total parameter count, while speed follows the smaller number of active parameters.
Questions
How much memory does DeepSeek V4 Pro need?
About 978 GB at Q4_K_M with an 8K context: 977 GB of weights, 0.1 GB of KV cache and about 1 GB of buffers. At 32K context it needs about 979 GB.
Can DeepSeek V4 Pro run on a Mac?
Not on a single Mac at Q4_K_M: it needs about 978 GB, more than the largest Mac memory size. It needs a cluster of machines or a lower-bit quantization.
How big is the DeepSeek V4 Pro KV cache?
About 0.1 GB at 8K tokens and 1.0 GB at 128K tokens with an F16 cache. Its attention design: compressed sparse attention.
Other models
- Gemma 4 12B
- gpt-oss 20B
- Gemma 4 26B-A4B
- Qwen3.8 27B
- GLM-4.7 Flash
- Gemma 4 31B
- Qwen3.6 35B-A3B
- gpt-oss 120B
- Mistral Medium 3.5
- Qwen3.8 Flash Next
- DeepSeek V4 Flash
- GLM-5.3
- Kimi K3
Data read from Hugging Face on 2026-11-09. These are estimates; benchmarks on this site report measured memory. How the estimate works. Compare all models in the model fit chart.
Get the next one
New benchmarks and agent runs by email, with the raw data. Or follow on RSS or X.