AI · October 21, 2026
Also published in Português (Brasil)
DeepSeek releases faster AI model
Chinese AI developer DeepSeek has released an improved version of its AI model that works faster and performs better in tasks such as programming and reasoning.
With the update, the company intensifies pressure on American competitors like OpenAI and Anthropic. DeepSeek aims to offer powerful AI at lower prices than the leading American players.
The new model, DeepSeek-V4.1-Flash-1017, was launched on Saturday. According to the Chinese company, it delivers better performance on benchmarks, executes tasks more quickly, and produces shorter responses than its predecessor. These improvements are expected to be useful for programming, building AI agents, and creating games and 3D models. The company describes the update as a relatively small enhancement of its earlier V4.1-Flash model.
DeepSeek has become a significant challenger to American AI companies, partly due to its low pricing strategy. While companies like OpenAI and Anthropic largely keep their most powerful models closed, DeepSeek offers models with open weights. Developers can use these weights to run or further customize the models themselves.
This pricing strategy also puts pressure on Chinese competitors. When DeepSeek introduced its slimmed-down V4.1-Flash model in early September, stock prices for companies like MiniMax, Z.ai, and Alibaba, which develops its own Qwen models, fell.
However, the rapid development has its downsides. Earlier this week, DeepSeek's servers were down for about five hours. Some experts suspect the company lacked sufficient computing power. There have also been frequent outages and delays in recent months.
Nonetheless, the competition continues, with DeepSeek emerging as a viable alternative to many American models. The latest model is available through DeepSeek's website and app, as well as via an API for developers.