AI

DeepSeek Launches V4.1-Flash-1017 With Open Search APIs

Saturday, October 17, 2026Read Original

Details

  • DeepSeek launched V4.1-Flash-1017, an updated model available through its website, app, and API.
  • The company says the release improves benchmark performance, speeds task completion, and produces shorter outputs than the previous V4.1-Flash.
  • The model uses reinforcement learning with self-feedback to support end-to-end project development, including game creation and 3D modeling.
  • It is optimized for DeepSeekHarness and Tencent CodeBuddy, extending its focus on software development and agentic workflows.
  • DeepSeek opened its Search and Fetch APIs to all users, with each endpoint priced at $0.75 per 1,000 successful calls.
  • The APIs use DeepSeek’s in-house search engine, which the company says handles hundreds of millions of requests daily worldwide.
  • DeepSeek released the model through Hugging Face and ModelScope, enabling external deployment and experimentation.
  • Published results show gains over the prior version, including Terminal-Bench 4.0 rising from 31.2 to 42.3 and ProgramBench from 20.3 to 29.1.

Impact

The release combines a stronger coding model with low-cost search and retrieval infrastructure, potentially making agentic development workflows more affordable for teams that need frequent tool calls. Its open distribution also pressures commercial model providers to compete on deployability and inference economics. However, benchmark comparisons remain dependent on each provider’s testing methodology, so real-world adoption will hinge on reliability, latency, licensing, and the quality of DeepSeek’s search results.

Rift Dispatch