DeepSeek V4 Models — Hardware Requirements
7 DeepSeek V4 models from DeepSeek and the community, from the smallest that runs in 14.6 GB of VRAM up to 889.5B parameters. Every row links to full quantization tables and GPU compatibility.
All DeepSeek V4 Models by Size
| Model | Params | Runs from | Context | Publisher | Quant downloads |
|---|---|---|---|---|---|
| DeepSeek V4 Flash JANG CRACK | 33.5B | 14.6 GB | 1049K | ||
| DeepSeek V4 Flash 162B | 92.2B | 39.5 GB | 1049K | ||
| DeepSeek V4 Flash 180B | 101.6B | 43.5 GB | 1049K | ||
| DeepSeek V4 Flash | 158.1B | 43.8 GB | 1049K | ||
| DeepSeek V4 Flash DSpark | 165.3B | 70.6 GB | 1049K | ||
| DeepSeek V4 Flash DSpark Abliterated Uncensored | 165.3B | 77.3 GB | — | ||
| DeepSeek V4 Pro | 861.6B | 366.5 GB | 1049K | ||
| DeepSeek V4 Pro DSpark | 889.5B | 378.4 GB | 1049K |
How DeepSeek V4 Compares — Benchmark Rating
DeepSeek V4 Pro is the highest-rated DeepSeek V4 model with an overall benchmark rating of 74.3/100 — #3 among 73 open models. The top proprietary model, Claude Fable 5 Max, scores 89.9. Click a model to see its full benchmark breakdown.
Claude Fable 5 Max · proprietary89.9
GPT 5.5 · proprietary89.2
GPT 5.6 Sol · proprietary89.2
Claude Fable 5 · proprietary88.6
Claude Opus 4.8 · proprietary88.1
GLM 5.282.7
Inkling79.2
DeepSeek V4 Pro74.3
Qwen3.6 27B74.0
Kimi K2.7 Code69.2
StableBeluga269.1
Frequently Asked Questions
- How much VRAM do I need to run a DeepSeek V4 model?
- The smallest DeepSeek V4 model, DeepSeek V4 Flash JANG CRACK, runs from 14.6 GB of VRAM at an aggressive quantization. Larger family members need proportionally more — see the table above for every model.
- Which DeepSeek V4 models can I run on a 16 GB GPU?
- 1 of 8 DeepSeek V4 models fit in 16 GB of VRAM at some quantization, including DeepSeek V4 Flash JANG CRACK.
- What is the most popular DeepSeek V4 model to run locally?
- DeepSeek V4 Flash is the most downloaded DeepSeek V4 model in local-friendly quantized formats. It runs from 43.8 GB of VRAM.
- How do DeepSeek V4 models score on benchmarks?
- DeepSeek V4 Pro leads the family with an overall benchmark rating of 74.3/100, ranking #3 among 73 open models, while the top proprietary model, Claude Fable 5 Max, scores 89.9. See the comparison chart above for the full standings.