DeepSeek
DeepSeek V2.5 236B
当前
8.9K下载量734点赞Sep 2024发布日期131K tokens上下文DeepSeek许可证80 优秀质量
DeepSeek V2.5 236B (236B parameters) requires approximately 204.1 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 21B 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 235 GB of VRAM.
快速开始
— 复制粘贴即可本地运行Copy-paste commands to run DeepSeek V2.5 236B on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "deepseek-ai/DeepSeek-V2.5" \
--hf-file "DeepSeek-V2.5-Q4_K_M.gguf" \
-c 4096 -ngl 99Quick specs
Parameters236B (21B active)
Architecturemoe (MoE)
Context131K tokens
Modalitytext
Min RAM92 GB
Rec. RAM144 GB (Q4_K_M)
LicenseDeepSeek
FamilyDeepSeek
✓ Chat✓ Reasoning
About this model
相关模型
快速推荐
最佳硬件
DeepSeek V2.5 236B 的最佳选择
运行此模型
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 | 92.0 GB | Low | — |
Q3_K_S | 3 | 115.6 GB | Low | — |
NVFP4 | 4 | 132.2 GB | Medium | — |
Q4_K_M | 4 | 144.0 GB | Medium | — |
Q5_K_M | 5 | 169.9 GB | High | — |
Q6_K | 6 | 193.5 GB | High | — |
Q8_0 | 8 | 252.5 GB | Very High | — |
F16 | 16 | 483.8 GB | Maximum | — |
Quality benchmarks
DeepSeek V2.5 236B benchmark scores
Reasoning
MMLU-Pro—
GPQA Diamond—
MATH-50082.8%
ARC Challenge—
Source: official · 2024-09-05
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
Weights144.0 GB
KV Cache58.6 GB
Runtime0.9 GB
Headroom0.6 GB
常见问题
FAQ — DeepSeek V2.5 236B
另请参阅