About the news
On 17 October 2026, DeepSeek made two related releases. First, it launched Search and Fetch APIs. Second, it updated the V4.1-Flash model (snapshot V4.1-Flash-1017), which remains available under the existing API name deepseek-flash.
What the Search and Fetch APIs do:
- Search API returns structured web results for a query
- Fetch API retrieves a single page and returns it as Markdown
- Both endpoints use the same authentication as the main DeepSeek API
- Pricing is listed at $0.75 per 1,000 successful requests
- Rate limit is 5 requests per second per account
- Certain server-side calls through supported harnesses were offered free for a limited period after launch
What the model update brings:
- The updated V4.1-Flash checkpoint is live on the website, app, and API
- DeepSeek reports stronger results on agentic and coding tasks, along with faster completion and shorter outputs compared with the previous Flash snapshot
- Open weights for the model family remain available under an MIT license
- The company positions the combination of the model and the new search tools as practical for agent builders and retrieval systems that need live web access at lower cost
Why it matters
The practical effect is straightforward. Developers building agents or RAG-style systems gain a cheaper coding model plus first-party web search and page retrieval behind one API key. This continues the pattern of Chinese labs competing on price and usable tooling for agent workloads.
Sources
- DeepSeek API documentation and changelog (October 17, 2026 updates)
- MarkTechPostwww.marktechpost.com/2026/10/17/deepseek-just-released-v4-1-flash-1017-scores-77-0-on-deepswe-opens-search-api/
