Inception tells customers it will shut down its Mercury diffusion-model API on 28 December

Inception, the startup behind the Mercury family of diffusion language models, has told customers it will wind down its Mercury API. In a notice emailed to API customers on 12 November, which AI Brief received as a customer, the company said the API "will remain fully available through December 28, 2026."

"After this date, you will no longer have access to the service or any data submitted to the platform," the notice says, urging customers to retrieve their data before then. It gives no reason for the decision and does not say what happens to the company or its models. Inception has not posted a public announcement.

Signs of the retreat were already visible. Inception's models page says Mercury 2 and Mercury 2.5 are no longer available to new users and that existing customers keep access until 15 November, even as it still advertises Mercury 3 at half price. Mercury 3 is no longer served through OpenRouter, according to a developer bug report filed on 12 November.

Inception's Mercury models generate text by diffusion, producing tokens in parallel rather than one at a time, and the company has marketed them on speed of more than 1,000 tokens per second.