DeepSeek V4.1 Flash preview: what changes for V4 Pro?

DeepSeek V4.1 Flash is planned for release around September 10, 2026, Beijing time (UTC+8), according to a screenshot of a usage-page notice supplied to Ofox. The notice also describes a temporary change for V4 Pro requests after Flash launches. This is a preview of that notice, not confirmation that the model is already live.

Source check: September 9, 2026. We could not independently retrieve the matching notice from a public DeepSeek page. The official pricing page checked for this article still listed V4 models, without a V4.1 Flash price table. The screenshot below is the evidence for the planned release and routing statements; the public documentation is a separate source for checking subsequent changes.

The supplied notice screenshot. It contains no visible page URL, exact launch time, numeric V4.1 prices or API model identifier.

What the notice says

The notice attributes its comparison to internal and external testing. It does not include benchmark tables, test settings or measurements. Treat the performance statement as a claim in the notice, not as an Ofox test result.

Why V4 Pro users should pay attention

If the announced routing takes effect, a request aimed at V4 Pro could be served by a different model during the transition. A stable client-side model string would not, by itself, prove that the serving model stayed unchanged.

Before the transition, save representative prompts, acceptance tests, outputs and billing records from your current workflow. After a confirmed change, compare the same tasks and inspect the provider’s usage records. For coding applications, include tool calls and structured results; a short chat response does not establish that the whole workflow is unchanged.

The notice names V4.1 Pro as the end of the temporary interval but gives no release date for it. It also does not say that existing V4 Flash requests will be changed in the same way. Do not extend the V4 Pro statement to every DeepSeek model.

Pricing and API access: what to wait for

Before committing a budget or editing a production configuration, confirm:

• The published API model ID and any alias or routing changes.
• Input, output and cache rates, including any peak/off-peak rules.
• The actual effective time and the account or endpoint covered.
• Supported parameters, limits and compatibility for the operations you use.

Do not turn the display name V4.1 Flash into a guessed API identifier. The current DeepSeek pricing documentation and change log are the places to check for published details. The existing DeepSeek pricing guide covers earlier V4 rates; those numbers are not a V4.1 quote.

If you use DeepSeek through Ofox

Check the Ofox model catalog for the actual entry, model ID and price available to your account. This preview does not announce V4.1 Flash availability on Ofox or promise that the routing in the screenshot applies to an Ofox request.

You can create an Ofox account to prepare your account setup, but registration does not grant advance access to an unlisted model. Wait for a confirmed catalog entry and billing terms before funding a V4.1-specific test. If you are evaluating API providers now, the OpenRouter migration guide explains which model, protocol and billing details to verify first.