DeepSeek V4.1 — DeepSeek 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
DeepSeek V4.1 — the Flash member shipped 10 September 2026 under MIT: a 552B-backbone MoE with a causal encoder-decoder design (8B active per input token, 16B per output token), a 196B-parameter Engram memory table and native image input. A V4.1 Pro was announced without a date and had not launched as of October 2026.
变体
DeepSeek V4.1 最小的变体在 Q4_K_M 下约需 462 GB 显存——量化权重加框架开销，不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| DeepSeek V4.1 Flash → 552B (16B active) | ~461.6 GB | 1,000,000 | 云端托管标签 | |
| DeepSeek V4.1 1.6T (49B active) | ~966.8 GB | 1,000,000 | ollama run deepseek-v4-1 |
显存为 Q4_K_M 下的量化权重加开销，与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 DeepSeek V4.1
安装 Ollama，然后拉取标签。
No local Ollama tag. Ollama lists deepseek-v4.1-flash only as a cloud-hosted ':cloud' tag, which runs on Ollama's servers, not your machine.
在上方选择一个尺寸，查看它自己的显存、速度估算和安装命令。
许可证
Commercial use permitted. No usage restrictions beyond attribution.
适用于: DeepSeek V4.1 Flash, DeepSeek V4.1我的 GPU 能运行 DeepSeek V4.1 吗？
- DeepSeek V4.1 在 AMD Ryzen AI Max+ 395 上
- DeepSeek V4.1 在 Apple M1 Ultra 上
- DeepSeek V4.1 在 Apple M2 Max 上
- DeepSeek V4.1 在 Apple M2 Ultra 上
- DeepSeek V4.1 在 Apple M4 Max 上
- DeepSeek V4.1 在 Apple M5 Max 上
- DeepSeek V4.1 在 NVIDIA A100 80GB (PCIe) 上
- DeepSeek V4.1 在 NVIDIA DGX Spark 上
- DeepSeek V4.1 在 NVIDIA H100 80GB (PCIe) 上
- DeepSeek V4.1 在 NVIDIA RTX PRO 6000 Blackwell 上
DeepSeek V4.1 — 常见问题
How much VRAM does DeepSeek V4.1 need?
DeepSeek V4.1 needs about 462 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: DeepSeek V4.1 Flash (462 GB, Q4_K_M); DeepSeek V4.1 (967 GB, Q4 (experimental, datacenter only)). On Apple Silicon, unified memory counts toward this requirement.
Can I run DeepSeek V4.1 on an RTX 4090 (24 GB)?
DeepSeek V4.1's smallest variant needs about 462 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for DeepSeek V4.1?
Q4_K_M is the best balance of quality and VRAM for DeepSeek V4.1 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run DeepSeek V4.1 with Ollama?
DeepSeek V4.1 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.