DeepSeek V3 Models — Hardware Requirements
9 DeepSeek V3 models from DeepSeek and the community, from the smallest that runs in 0.3 GB of VRAM up to 685.4B parameters. Every row links to full quantization tables and GPU compatibility.
All DeepSeek V3 Models by Size
| Model | Params | Runs from | Context | Publisher | Quant downloads |
|---|---|---|---|---|---|
| DeepSeek v3 0324 | 684.5B | 294.8 GB | 164K | ||
| DeepSeek V3.1 | 684.5B | 192.1 GB | 164K | ||
| DeepSeek V3.1 Terminus | 684.5B | 192.1 GB | 164K | ||
| DeepSeek v3 | 684.5B | 294.8 GB | 164K | ||
| DeepSeek V3.2 | 685.4B | 192.3 GB | 164K | ||
| DeepSeek V3.2 Exp | 685.4B | 295.2 GB | 164K | ||
| DeepSeek V3.2 Speciale | 685.4B | 295.2 GB | 164K |
How DeepSeek V3 Compares — Benchmark Rating
DeepSeek V3.1 is the highest-rated DeepSeek V3 model with an overall benchmark rating of 55.5/100 — #17 among 78 open models. The top proprietary model, Claude Fable 5 (high), scores 87.8. Click a model to see its full benchmark breakdown.
Claude Fable 5 (high) · proprietary87.8
Claude Opus 4.7 · proprietary86.4
Claude Fable 5 Max · proprietary82.4
GPT 6 Astra (high) · proprietary82.4
GPT 6 Astra Max · proprietary82.4
Kimi K366.8
GLM 565.9
DeepSeek R1 052863.1
DeepSeek V3.155.5
DeepSeek v3 032451.6
DeepSeek V3.246.2
DeepSeek v343.5
Frequently Asked Questions
- How much VRAM do I need to run a DeepSeek V3 model?
- The smallest DeepSeek V3 model, DeepSeek V3.1, runs from 192.1 GB of VRAM at an aggressive quantization. Larger family members need proportionally more — see the table above for every model.
- Which DeepSeek V3 models can I run on a 16 GB GPU?
- No DeepSeek V3 model currently fits in 16 GB of VRAM — the family starts at 0.3 GB.
- What is the most popular DeepSeek V3 model to run locally?
- DeepSeek v3 0324 is the most downloaded DeepSeek V3 model in local-friendly quantized formats. It runs from 294.8 GB of VRAM.
- How do DeepSeek V3 models score on benchmarks?
- DeepSeek V3.1 leads the family with an overall benchmark rating of 55.5/100, ranking #17 among 78 open models, while the top proprietary model, Claude Fable 5 (high), scores 87.8. See the comparison chart above for the full standings.