we_are_coded.by CODE · The world, decoded
БГ

DeepSeek · Open weights

DeepSeek opened its search to everyone: $0.75 per 1,000 calls, free through Claude Code and Codex until October 31

On October 17 DeepSeek released V4.1-Flash-1017, a minor update to its model with shorter outputs and weights under MIT. The big item in the announcement is the Search & Fetch API: $0.75 per 1,000 calls per endpoint, free through DeepSeek Harness, Claude Code and Codex until billing starts on October 31 at 16:00 UTC.

Source
DeepSeek API Docs
Published
The event
Checked on
Responsible editor
Cvetelin Ivanov
How we work
MethodCorrections

The model card calls it a minor update. The banner at the top of DeepSeek's docs is about something else: search.

Fact

The facts: on October 17, 2026 DeepSeek released DeepSeek-V4.1-Flash-1017. Its Hugging Face card describes it as a minor update to DeepSeek-V4.1-Flash and promises stronger results, faster task completion and shorter outputs; the model is live on the website, the app and the API. The context is 1 million tokens, the recommended max_tokens is at least 256K, and the weights are under the MIT license. The DeepSeek Search & Fetch API is open to everyone; by DeepSeek's account it runs on an in-house search engine serving hundreds of millions of requests daily worldwide. Each endpoint costs $0.75 per 1,000 calls. Calls through server-side tools in DeepSeek Harness, Claude Code and Codex are currently free for DeepSeek API users; billing starts October 31, 2026 at 16:00 UTC. By DeepSeek's numbers in the card's table: DeepSWE v1.1 - 77.0 against 74.2 for the previous Flash, Terminal-Bench 4.0 - 42.3 against 31.2; in the same table Opus 5 is ahead of it in most rows, on Terminal-Bench 4.0 with 51.8. The announcement gives no price for the model itself.

The price is per endpoint. Search is billed separately from fetching a page. An agent that does both pays for both.

The free period applies exactly to server-side tools in DeepSeek Harness, Claude Code and Codex. Two of the three aren't DeepSeek's: Claude Code and Codex.

The model is an update. The product is the search.

Hundreds of millions of requests a day is DeepSeek's claim, with no outside check. The results table is its own too. In it the new Flash leads on DeepSWE v1.1, but Opus 5 is ahead in most rows.

The backstoryEarlier DeepSeek released V4.1-Flash under MIT and redirected requests from its larger model to it.

Before October 31 at 16:00 UTC, try the search through Claude Code or Codex and count how many calls your agent burns. After that date they're on your bill.

Official primary sources

DeepSeek API Docs - DeepSeek-V4.1-Flash-1017 Release, 17.10.2026Hugging Face - DeepSeek-V4.1-Flash-1017 model card (MIT)

The visual is generated code-art, no third-party images.

Original: https://wearecoded.com/en/articles/deepseek-v41-flash-1017-search-fetch.html

More on this

Frontier
FrontierMiniMax · 25 October 2026MiniMax released M3.1-Flash: a multimodal model with about 428 billion parameters, 23 billion active and a context of up to 1 million tokensFrontierDeepSeek API Docs · 10 September 2026DeepSeek released V4.1-Flash and retired its own flagship: requests to V4-Pro now go to the smaller modelFrontierHugging Face · 15 October 2026Meta opened Muse Spark Open at 816 billion parameters: free, but with a “Built with Muse Spark Open” badge and a contract with Meta if you sell it as a service with revenue above 20 million dollars
Back to all news