DeepSeek V4.1 Flash 1017: Open-Weight Update Cuts Agent Costs, Search API Opens to All

On Saturday, October 17, 2026, DeepSeek quietly upgraded its fast model. According to the DeepSeek API changelog, V4.1 Flash is now served by DeepSeek-V4.1-Flash-1017 on the web, in the app and through the API. Pricing and the calling method don’t change: developers keep using the deepseek-flash model name.

What’s new in V4.1 Flash 1017

DeepSeek calls it a “minor update”, but the Hugging Face model card lists real gains in reasoning, coding, agent and vision tasks, with shorter outputs and faster task completion. The weights are open under the MIT license, with a context window of up to 1 million tokens.

• DeepSWE v1.1: 77.0, up from 74.2 (Claude Opus 5: 74.0, GPT-5.6 Sol: 73.0)
• Terminal-Bench 4.0: 42.3, up from 31.2
• Automation-Bench: 60.5, up from 54.8

These are DeepSeek’s own numbers. Its table also shows Opus 5 still ahead on several tests, such as Humanity’s Last Exam (56.3 vs 42.6) and Terminal-Bench 4.0 (51.8).

The real story: cost per task

As The Decoder points out, the shorter answers matter most. On DeepSeek’s DeepSWE cost chart, the new model reaches 77% at just under $0.20 per task, and about 72% for roughly $0.10 in its low setting. The previous version cost around $0.50 per run. Independent benchmarks are not out yet.

DeepSeek opens its search engine

DeepSeek is also opening its in-house Search and Fetch APIs to all developers. The company says its AI-native search engine indexes hundreds of billions of web pages, handles hundreds of millions of requests a day and filters out SEO spam and low-quality AI content. Those are vendor claims. Investing.com confirmed the Saturday release.

Rocks take

For anyone running AI agents, coding assistants or Home Assistant automations through an API, price per finished task is what counts, and this is where DeepSeek keeps pushing. The weights are open, but this is a very large model: running it at home is not realistic for most of us. The interesting part for makers is the cheap API plus a built-in search tool.