DeepSeek updates V4.1 Flash with claimed speed and benchmark gains

DeepSeek released V4.1-Flash-1017 on October 17, updating the model available through its website, app and API. The company said the version improves benchmark results while completing tasks faster and producing less text.

DeepSeek listed a Terminal-Bench 4.0 score of 42.3, compared with 31.2 in its September release results. ProgramBench rose from 20.3 to 29.1. These figures are company-published evaluations, with the latest Terminal-Bench test conducted using Claude Code.

The update follows the September 10 introduction of V4.1 Flash, which supports visual understanding alongside text. DeepSeek said the latest changes reduce task costs through shorter outputs and faster execution. API rates remain unchanged.

Developers can access the updated model through the existing deepseek-flash identifier. DeepSeek said it optimized the release for DeepSeek Harness, Tencent CodeBuddy and WorkBuddy. It recommends Harness version 0.2.1-rc.1 or newer.