DeepSeek releases AI model update, opens search APIs to all users
Chinese artificial intelligence firm DeepSeek unveiled an enhanced version of its AI model on Saturday, DeepSeek-V4.1-Flash-1017. The company highlighted the new model's superior benchmark performance, increased speed, and condensed output lengths in its announcement. DeepSeek-V4.1-Flash-1017 is now accessible via the company's website, mobile application, and application programming interface (API), as reported on its official X account.
The update from DeepSeek was characterized as a "Pareto improvement," signifying gains in functionalities and efficiency while avoiding any rise in costs. This model has been optimized for DeepSeek Harness, Tencent CodeBuddy, WorkBuddy, and other tools. Additionally, the company introduced its Search and Fetch APIs to every user, powered by their internal search engine. Each Search or Fetch API endpoint is priced at $0.75 for every 1,000 calls, according to DeepSeek.
DeepSeek-V4.1-Flash-1017 demonstrated notable improvements over its predecessor, achieving a score of 77.0 on the DeepSWE v1.1 benchmark, an increase from the previous 74.2 score. It also performed better on Automation-Bench V1.0.6, scoring 60.5, up from the prior 54.8.
Written by urgent.news from Investing.com's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
