China’s DeepSeek Launches V4.1-Flash Model
DeepSeek says the smaller model improves coding and agent tasks while cutting prices by as much as 32%, Reuters reported.
- On Thursday, Chinese artificial intelligence startup DeepSeek launched DeepSeek-V4.1-Flash, the smallest model in its new architecture family, designed for greater capability and faster inference.
- Pre-Trained on 45 trillion tokens, V4.1-Flash handles up to 1 million tokens of context and understands images natively while activating 8 billion parameters for input and 16 billion for output.
- Performance tests show the model scores 88.1 on CyberGym and 74.2 on DeepSWE, outperforming Anthropic's Claude Opus and OpenAI's GPT-5 Sol, while off-peak pricing is $0.60 per million tokens.
- Until V4.1-Pro arrives, DeepSeek is automatically routing all V4-Pro customer requests to the new architecture and billing them at cheaper Flash rates, reversing August price increases.
- As the company prepares for an IPO on Shanghai's STAR Market, shares in rivals MiniMax and Z.ai fell more than 8% in Hong Kong on Thursday, with Alibaba slipping more than 2%.
30 Articles
30 Articles
DeepSeek has started another price war in the artificial intelligence market, the Chinese company's new model offers performance close to that of its American and Chinese rivals at a fraction of the price. V4.1 Flash could cause a stir, especially for the increasingly popular AI agents that work independently for long periods of time, so cost is a key issue when using them. The announcement also caused serious waves on the stock market, with the…
DeepSeek's V4.1-Flash takes on Opus 5, and retires V4-Pro
DeepSeek released V4.1-Flash on Thursday. It is a smaller model that the company says beats its own flagship on coding and agent tasks, at a lower price. The Hangzhou lab announced it in a thread on X and published the weights on Hugging Face under the MIT licence. That lets anyone download, modify and run […] This story continues at The Next Web
DeepSeek introduced V4.1 Flash, which exceeds V4-Pro in coding and task solving using agents.
Coverage Details
Bias Distribution
- 72% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
