DeepSeek-V4.1-Flash changelog

Vals Index: Muse Wonder Full fourth, DeepSeek V4.1 Flash 1017 eighth at $0.66 per test

The Vals Index, updated October 22, 2026, ranks Muse Wonder Full fourth at 66.79% and Muse Wonder sixth at 64.88%. DeepSeek V4.1 Flash 1017 is eighth at 61.26% for $0.66 per test, GLM 5.3 ninth at 59.73%, Grok 4.7 12th at 58.56% and Kimi K3 19th at 52.95%.

DHS orders US federal civilian agencies to remove DeepSeek products

A Department of Homeland Security order published in the Federal Register on October 20, 2026 tells federal civilian agencies to stop buying DeepSeek products, including third-party products built with DeepSeek technology, and remove them from their systems. Agencies can request exceptions within 60 days.

Who is affected:

US federal civilian agencies and vendors that sell them products built on DeepSeek models.

Action needed:

If you sell to US federal civilian agencies, check whether your product uses DeepSeek models.

Epoch: DeepSeek-V4.1-Flash-1017 scores 93% on GPQA Diamond and 71% on FrontierMath tiers 1-3

Epoch benchmark data, with runs started October 19, 2026, records deepseek-v4.1-flash-1017 at max effort at 0.932 on GPQA Diamond, 0.712 on FrontierMath tiers 1-3 and 0.463 on tier 4.

DeepSeek starts billing server-side search and fetch calls October 31

Search and Fetch API calls cost $0.75 per 1,000 requests. Calls made through server-side tools in DeepSeek Harness, Claude Code and Codex are free until billing starts at 16:00 UTC on October 31, 2026.

Who is affected:

Developers using DeepSeek models in Claude Code, Codex or DeepSeek Harness with web search.

What to do:

Expect new per-request charges for search and fetch after October 31, 2026.

DeepSeek-V4.1-Flash-1017 update released with open weights

DeepSeek released DeepSeek-V4.1-Flash-1017 on its website, app and API with stronger benchmarks, faster task completion and shorter outputs, and published the weights on Hugging Face. deepseek-flash now serves this version.

Who is affected:

Apps calling deepseek-flash, which now get the 1017 version.

What to do:

Retest prompts that depend on output length or style.

DeepSeek-V4.1-Flash in public preview on Gemini Enterprise Agent Platform

DeepSeek-V4.1-Flash is available as a fully managed model in Model Garden on Gemini Enterprise Agent Platform, in public preview, with text and image input and a context window of up to 1M tokens.

DeepSeek lowers API prices with V4.1-Flash

DeepSeek lowered API prices with the V4.1-Flash launch, keeping off-peak rates at 50% of peak, effective 04:00 UTC on September 10, 2026. The pricing page lists deepseek-flash at $0.30 input (cache miss) and $1.20 output per 1M tokens at peak.

DeepSeek-V4.1-Flash replaces V4-Flash, V4-Flash retired

DeepSeek released V4.1-Flash with native vision under the model name deepseek-flash, and retired V4-Flash and V4-Flash-Vision-Exp. The old names temporarily route to V4.1-Flash.

Who is affected:

Apps calling deepseek-v4-flash or deepseek-v4-flash-vision-exp.

Action needed:

Set the model name to deepseek-flash.