DeepSeek
DeepSeek V3.1 671B
前沿
544.3K下载量3.1K点赞Mar 2025发布日期131K tokens上下文DeepSeek许可证83 优秀质量
DeepSeek V3.1 671B (671B parameters) requires approximately 470.4 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 37B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 541 GB of VRAM.
快速开始
— 复制粘贴即可本地运行Copy-paste commands to run DeepSeek V3.1 671B on your machine.
Run
ollama run deepseek-v3:0324Quick specs
Parameters671B (37B active)
Architecturemoe (MoE)
Context131K tokens
Modalitytext
Min RAM261.7 GB
Rec. RAM409.3 GB (Q4_K_M)
LicenseDeepSeek
FamilyDeepSeek
✓ Code✓ Chat✓ Reasoning
About this model
- •Major improvements in instruction following and structured output generation
- •Stronger coding performance — competitive with frontier proprietary models
- •Enhanced agentic and tool-use capabilities
- •Same 671B MoE architecture (37B active) as V3 — no additional VRAM needed
相关模型
量化选项
各量化级别的 VRAM 估算
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 261.7 GB | Low | — |
Q3_K_S | 3 | 328.8 GB | Low | — |
NVFP4 | 4 | 375.8 GB | Medium | — |
Q4_K_M | 4 | 409.3 GB | Medium | — |
Q5_K_M | 5 | 483.1 GB | High | — |
Q6_K | 6 | 550.2 GB | High | — |
Q8_0 | 8 | 718.0 GB | Very High | — |
F16 | 16 | 1375.6 GB | Maximum | — |
Quality benchmarks
DeepSeek V3.1 671B benchmark scores
Coding
SWE-bench Verified49.2%
HumanEval+85.3%
Aider Polyglot56.9%
LiveCodeBench56.4%
Reasoning
MMLU-Pro78.1%
GPQA Diamond63.0%
MATH-50092.4%
ARC Challenge—
General
Chatbot Arena—
IFEval88.5%
Source: official · 2025-03-24
硬件兼容性
全部硬件的适配估算
Computing compatibility...
内存详细分析
Reference: RTX 2060 6GB
Weights409.3 GB
KV Cache59.6 GB
Runtime0.9 GB
Headroom0.6 GB
常见问题
FAQ — DeepSeek V3.1 671B
另请参阅