DeepSeek Releases V4-Pro, an MIT-Licensed MoE Model
The company's newest flagship targets reasoning and coding while keeping a permissive open-source license.
DeepSeek has published DeepSeek-V4-Pro-0813, the high-end entry in its new V4 family, on Hugging Face. The model uses a mixture-of-experts (MoE) architecture and is aimed at demanding reasoning and coding workloads.
The release carries an MIT license, one of the most permissive terms available. That means developers can use, modify, and deploy the weights commercially with minimal restrictions — a notable stance for a frontier-class model and consistent with DeepSeek's earlier open releases.
Why it matters
MoE designs activate only a fraction of their parameters per token, letting labs scale total capacity while holding inference costs down. Pairing that efficiency with a genuinely open license is what has made DeepSeek's models a fixture in the open-weights ecosystem.
Key points from the record:
- Family: V4-Pro, positioned as the flagship tier
- Focus: text generation, reasoning, and code
- Architecture: mixture-of-experts
- License: MIT
DeepSeek has not published detailed specifications such as parameter counts or context length alongside this listing, so builders will want to consult the model card directly as further documentation and benchmarks appear.
Sources
More from DeepSeek
All DeepSeek releases →DeepSeek Releases V4.1-Flash, an Open Multimodal MoE
The new mixture-of-experts model handles text, vision, and reasoning, and ships under a permissive MIT license.
DeepSeek V4.1-Flash Adds Vision and Reasoning
The new MIT-licensed checkpoint folds image understanding and reasoning into DeepSeek's fast, mixture-of-experts Flash line.
DeepSeek adds vision to its V4 Flash line
An experimental, MIT-licensed vision-language model brings image understanding to DeepSeek's fast V4 Flash architecture.
More in Text / LLM
All Text / LLM →
Microsoft's Aion 1.0 Instruct is a compact 2.3B chat model
A small, multilingual instruction-tuned model aimed at chat and summarization lands on Hugging Face.
SenseTime Releases Looped 9B Multimodal Model
The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Tencent releases Hunyuan Hy4-flash multimodal MoE
The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.
0 comments
No comments yet. Be the first to weigh in.