US EditionDeepSeek's New V3.1 Hints at Potent New Chinese Chips
DeepSeek's AI model upgrade supports next-generation domestic chips with a hybrid inference structure and will lower API costs from September 6, promoting wider adoption across sectors.
- On August 21, 2025, in Beijing, Chinese AI startup DeepSeek launched V3.1, a new iteration of its leading AI model.
- The upgrade addresses U.S. semiconductor export restrictions by optimizing the model for soon-to-be-released Chinese-made chips to reduce reliance on foreign technology.
- DeepSeek V3.1 features faster processing speeds, enhanced reasoning and non-reasoning capabilities via a hybrid inference structure, and supports an expanded context window of 131,072 tokens.
- The company reported that V3.1 scored 30 on the Browsecomp benchmark, a notable improvement over the May R1 update's 8.9, and emphasized enhancements in its ability to utilize external tools and execute function calls, which are essential for agentic AI tasks.
- DeepSeek plans to reduce API pricing from September 6 to improve developer access, signaling a push for wider adoption and positioning as a competitive global AI provider.
Insights by Ground AI
Does this summary seem wrong?
24 Articles
24 Articles
Chinese AI startup DeepSeek launches enhanced V3 AI model with China-made chip support and faster speeds - Tech Startups
Chinese artificial intelligence startup DeepSeek has rolled out an upgraded version of its flagship model, underscoring the country’s broader effort to cut reliance on U.S. technology. The new release, called DeepSeek-V3.1, adds support for China-made chips and promises faster processing […] The post Chinese AI startup DeepSeek launches enhanced V3 AI model with China-made chip support and faster speeds first appeared on Tech Startups.
After months of silence Deepseek presents a new model. But even in the People's Republic, the reaction is initially subdued.
·Frankfurt, Germany
Read Full ArticleCoverage Details
Total News Sources24
Leaning Left5Leaning Right4Center1Last UpdatedBias Distribution50% Left
Bias Distribution
- 50% of the sources lean Left
50% Left
L 50%
R 40%
Factuality
To view factuality data please Upgrade to Premium
