Chinese AI company Moonshot AI is targeting $2 billion in annualized revenue by the end of 2026, twice its reported run rate in August, according to Bloomberg. The target follows growing adoption of Kimi K3, the company’s open-weight AI model released this summer.
The planned increase would represent rapid commercial growth for one of China’s most prominent AI labs. K3 usage has declined slightly from earlier peaks, but OpenRouter data has shown K3 models generating as many as 300 billion tokens per day on its platform.
Kimi K3 Drives Moonshot’s Growth
Moonshot introduced K3 in July as its most capable model for coding, knowledge work and reasoning. In its official Kimi K3 announcement, the company also committed to releasing the model weights, allowing developers and companies to deploy the model outside Moonshot’s own services.
That open-weight approach differs from companies such as OpenAI and Anthropic, which keep their leading model weights private and primarily charge users for access through subscriptions and APIs. Moonshot’s projected revenue remains substantially below its U.S. competitors, with recent reports putting OpenAI’s annualized revenue at about $40 billion and Anthropic’s at roughly $65 billion.
Freely available weights also mean customers can run K3 independently rather than paying Moonshot for every interaction. Even so, the company’s rising revenue target suggests it is generating substantial sales around an open-weight model through its own products and services.
Anthropic Alleges Large-Scale Distillation Campaign
Moonshot’s model-development practices have also come under scrutiny. Anthropic alleged in its September threat intelligence report that Moonshot conducted a large unauthorized distillation campaign against Claude between May and July 2026.
Anthropic said it observed more than 23 million exchanges linked to the campaign. During one 10-day period, it said Moonshot routed almost 300,000 customer requests through a network of 5,380 fraudulent accounts, with most requests sent to Claude’s Opus models.
According to Anthropic, some users believed they were interacting with Kimi when their requests were instead forwarded to Claude. The company also alleged that Moonshot saved some of those exchanges and developed techniques to extract Claude’s reasoning traces for model training.
Anthropic said some forwarded conversations contained sensitive customer information, including internal company data and surveillance material. The report did not include a response from Moonshot to the allegations.
Featured image credits: SearchYour.ai
For more stories like it, click the +Follow button at the top of this page to follow us.