Archive — CN — technode
DeepSeek open-sources new V3.1 model
DeepSeek open-sourced its V3.1-Base model with 685 billion parameters and an extended context length of 128K on August 20. The upgraded model is available for testing on the company's platforms, but the release date for the upcoming DeepSeek R2 model remains unconfirmed.
Key points
- DeepSeek announced the open-sourcing of its V3.1-Base model on August 20.
- The new model features approximately 685 billion parameters.
- The context length has been extended to 128K, improving its processing capability.
- Users were notified of the upgrade to V3.1 earlier in the evening on the same day.
- The model can be tested on the official website, app, and mini-program without changes to the API call method.
- There is currently no confirmed release date for the anticipated DeepSeek R2 model.
Why it matters
- The trend of open-sourcing AI models aligns with the growing demand for transparency and collaboration in AI development.
- As AI models become more complex, the increase in parameters and context length reflects the industry trend towards developing more advanced language processing capabilities, which is a key aspect of AI in 2025.
The open-sourcing of DeepSeek's V3.1 model marks a significant evolution in AI capabilities, reflecting ongoing trends in transparency and complexity within the industry.