As competition in the field of artificial intelligence accelerates, China-based DeepSeek has made a new update to its model family. The company announced that the DeepSeek-V4.1-Flash-1017 version is now accessible via its website, mobile application, and API.
In a statement made on DeepSeek’s official X account, it was noted that the new version provides improvements in performance and efficiency without increasing costs. The company described this approach as a “Pareto improvement.”
According to the shared benchmark results, the model scored 77.0 in the DeepSWE v1.1 test and 60.5 in the Automation-Bench V1.0.6 test. These results indicate that a stronger performance is targeted, particularly in software development and automation-focused tasks.
SEARCH AND RETRIEVAL APIS OPENED
One of the highlights of the update was that DeepSeek made its search and retrieval APIs, supported by its own search engine, available to all developers. This will allow developers to integrate the DeepSeek infrastructure more directly into their applications.
According to the company's announcement, a fee of 0.75 dollars per 1,000 calls will be applied for each of the search and retrieval API endpoints. This pricing offers a more predictable model for developers who want to calculate costs based on usage volume.
It is stated that DeepSeek-V4.1-Flash is supported by a self-feedback reinforcement learning infrastructure. This structure is expected to increase the model's output quality and task completion efficiency in more comprehensive project development processes.
The company states that the new model can also provide more effective results in end-to-end development tasks such as game creation and generating 3D models from scratch. These areas are among the topics where generative AI is increasingly merging with software and design processes.
DeepSeek’s latest move shows that in AI services, not only model capabilities but also the balance of accessibility and cost are becoming increasingly decisive. The new API opening reveals that the company aims for wider usage in the developer ecosystem.
News Source: 12punto
