DeepSeekNews

DeepSeek ships V4.1-Flash-1017 open weights and opens its Search & Fetch APIs

DeepSeek upgraded V4.1 Flash to V4.1-Flash-1017 (MIT-licensed weights; API price unchanged) and opened its in-house web search engine to developers as Search and Fetch APIs at $0.75 per 1,000 successful calls.

DeepSeek-V4.1-Flash-1017 is a minor update to V4.1 Flash. It now powers the DeepSeek web app, mobile app and API.

  • API: keep calling deepseek-flash. DeepSeek says the calling method and pricing are unchanged.
  • Benchmarks (DeepSeek’s own numbers): DeepSWE v1.1 77.0 (up from 74.2) and Terminal-Bench 4.0 42.3 (up from 31.2). DeepSeek also says tasks finish faster with shorter outputs.
  • Open weights: on Hugging Face under the MIT License, with a 1M-token context window.

New Search & Fetch APIs: DeepSeek opened its in-house search engine to developers.

  • POST /v1/tools/search returns ranked web results, with optional page content.
  • POST /v1/tools/fetch returns a single page as Markdown.
  • Price: $0.75 per 1,000 successful requests for each API. Failed requests, empty results and timeouts aren’t billed. The limit is 5 requests per second per account for each API.
  • Server-side search calls made through DeepSeek Harness, Claude Code or Codex are free until Nov. 1, 2026 (Beijing time). Billing starts then.

Why it matters

It's a cheap, strong open model, and DeepSeek says it now gets better results with shorter outputs. A pay-per-call search API also gives agents built-in web grounding without a separate search provider.

Sources