AI News · Models ·
DeepSeek updates V4.1-Flash for faster agent tasks and opens a Search API
DeepSeek released V4.1-Flash-1017, a minor update it says cuts task execution time and output length, and reported 77% on the DeepSWE v1.1 coding test, up from 74.2%. It also opened paid Search and Fetch APIs at 75 cents per 1,000 successful requests, while API prices for the model stay unchanged.
Key points
- V4.1-Flash-1017 scored 77% on DeepSWE v1.1 in DeepSeek's tests, up from 74.2%.
- New Search and Fetch APIs cost 75 cents per 1,000 successful requests.
- Weights are on Hugging Face under the MIT license, with API prices unchanged.
What happened: DeepSeek released DeepSeek-V4.1-Flash-1017, an update to the V4.1-Flash model it launched on September 10, SiliconANGLE reported. The Hugging Face model card calls it a "minor update" that "reduces task execution time and output length." DeepSeek did not say by how much.
The details: In DeepSeek's own testing, the update resolved 77% of tasks on the DeepSWE v1.1 software engineering benchmark, up from 74.2% for the September version. Anthropic's Claude Opus 5 scores 74% in the same table, though Claude still leads on seven of the 11 benchmarks DeepSeek published. The company said it tuned the model for agent software, including its DeepSeek Harness tool and Tencent's CodeBuddy and WorkBuddy products. The model is live in DeepSeek's apps, available through the existing deepseek-flash API model ID at unchanged prices, and its weights are on Hugging Face under the MIT license.
The details: DeepSeek also opened its in-house search engine to developers. A Search API returns up to 50 ranked web results per query, and a separate Fetch API converts a web page to Markdown. Both cost 75 cents per 1,000 successful requests, and calls made through DeepSeek Harness, Claude Code or OpenAI's Codex are free until November 1.
What to watch: The releases follow a nearly five-hour API outage on October 15. DeepSeek apologized, gave users credits matching their previous week's spend and said it would remove single points of dependency in its systems.
Our take
Teams running coding agents on a budget get a cheaper open-weight option to test, but should weigh DeepSeek's recent five-hour API outage.