Moonshot AI sets $2B annual revenue target for K3 model amid open-weight debate

Moonshot AI, the Chinese AI lab behind the open-weight Kimi models, announced a $2 billion annual revenue target for 2026, doubling its reported run rate for August. The goal reflects the company’s push to monetize its K3 model, which has seen significant usage since its release earlier this summer.

OpenRouter data cited in the report indicates that K3 models currently generate up to 300 billion tokens per day on the platform, though the article notes that K3’s usage figures have declined slightly in recent months. Despite this traction, Moonshot’s projected revenue remains far below that of its closed-weight competitors: OpenAI and Anthropic, which recent reports put at $40 billion and $65 billion in annual revenue, respectively.

The company’s reliance on open-weight models introduces margin pressures compared to closed-weight frontier models, as model weights are freely available. This dynamic has not deterred Moonshot from pursuing aggressive growth targets, signaling confidence in the commercial potential of open-weight AI systems.

Moonshot’s development practices have come under scrutiny. Anthropic accused the company of a long-running model distillation campaign, alleging that nearly 300,000 requests from Kimi were routed to Claude Opus, effectively serving Opus in place of Kimi’s own models. Anthropic further claimed that more than 23 million responses were collected from its models for use in Moonshot’s training, raising questions about the integrity of Moonshot’s model development pipeline.