Chinese AI startup DeepSeek releases upgraded model with domestic chip support
Summary
Chinese AI startup DeepSeek launched an upgraded V3.1 model optimized for upcoming domestic Chinese chips, featuring faster processing and a new hybrid inference structure.
Key Points
- DeepSeek released an upgraded V3.1 AI model optimized for soon-to-be-released domestic Chinese chips.
- The V3.1 model uses FP8 precision format to enhance efficiency and speed.
- The upgrade introduces a hybrid inference structure allowing operation in reasoning and non-reasoning modes.
- DeepSeek will adjust API usage costs for the model starting September 6.
