全球开源模型频道

超大 MoE；机房多卡 / 云推理，不适合单消费卡。

R1 蒸馏小模型；消费级可跑推理风格。

ollama run deepseek-r1:8b

蒸馏 32B；24GB 卡推理能力明显强于 8B。

ollama run deepseek-r1:32b

代码 MoE；Lite 变体更易本地。

DeepSeek Coder is a capable coding model trained on two trillion code and natural language tokens.

ollama run deepseek-coder

An advanced language model crafted with 2 trillion bilingual tokens.

ollama run deepseek-llm

A strong, economical, and efficient Mixture-of-Experts language model.

ollama run deepseek-v2

DeepSeek-OCR is a vision-language model that can perform token-efficient OCR.

ollama run deepseek-ocr

An upgraded version of DeekSeek-V2 that integrates the general and coding abilities of both DeepSeek-V2-Chat and DeepSeek-Coder-V2-Instruct.

ollama run deepseek-v2.5

DeepSeek-V3.1-Terminus is a hybrid model that supports both thinking mode and non-thinking mode.

ollama run deepseek-v3.1

DeepSeek-V4-Flash is a preview of the DeepSeek-V4 series, a Mixture-of-Experts model with 284B total parameters and 13B activated, built for efficient reasoning across a 1M-token context window.

ollama run deepseek-v4-flash

DeepSeek-V4-Pro is a frontier Mixture-of-Experts model with a large context window and three reasoning modes.

ollama run deepseek-v4-pro

Hugging Face GGUF · downloads ~712,351. Repo: antirez/deepseek-v4-gguf

Hugging Face GGUF · downloads ~310,094. Repo: huihui-ai/Huihui-DeepSeek-V4-Flash-abliterated-ds4-GGUF