Moonshot AI halts new Kimi K3 signups as demand hits GPU limits

Three days after launching Kimi K3 — the largest open-weight AI model ever released — Chinese startup Moonshot AI suspended new consumer subscriptions on Sunday, July 19, saying user demand had pushed its computing infrastructure to its limits.

"To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members," the company wrote on X. Moonshot AI added that existing subscribers would not be affected and that it was "adding capacity as fast as we can," with plans to reopen new subscription spots in batches.x

A Victim of Its Own Success

Moonshot AI launched Kimi K3 on July 16, introducing a 2.8 trillion-parameter mixture-of-experts model with a one-million-token context window and native vision capabilities. The model is the first open-weight system to approach the 3-trillion-parameter mark, dwarfing DeepSeek V4's 1.6 trillion parameters.bbc+2

According to the company, user request volume over the first 48 hours far exceeded projections, approaching the upper limit of its GPU computing capacity. In addition to the subscription pause, Moonshot AI announced it would split its membership system into two products — "Kimi Membership" for the web platform, app, and office use cases, and "Kimi Code Membership" for programming workflows — to allocate computing resources more efficiently.youtube+1

Closing the Frontier Gap

The surge in demand reflects the attention Kimi K3 has drawn for its performance against leading U.S. models. Independent evaluations placed K3 close behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on overall benchmarks, while outperforming both on tasks like front-end coding and long-context work, according to Reuters and CNBC. The model scored 57 on Artificial Analysis's Intelligence Index, trailing only Fable at 60 and Sol at 59.ajscloud+2

K3's API is priced at $3 per million input tokens and $15 per million output tokens — matching Anthropic's Claude 5 Sonnet tier but far above pricing from other Chinese labs. The full model weights remain on track for open-source release by July 27, a move that could shift demand pressure away from Moonshot's own infrastructure and onto the broader developer ecosystem.servergurus+3