Kimi K3 Halts New Signups as Demand Overwhelms Moonshot's Compute

Moonshot AI paused new consumer subscriptions to Kimi K3 on July 19, saying it was “temporarily pausing new subscriptions and prioritizing compute for current members.” The 2.8-trillion-parameter mixture-of-experts model — with a one-million-token context window and native vision — launched on July 16 and days later topped Arena’s front-end coding leaderboard globally with 1,679 points. The surge simply outran the company’s GPU capacity. Existing subscribers were unaffected, and Moonshot said it would reopen new spots in batches. It is the second high-profile capacity crunch of the month, a reminder that for the most capable agentic models, serving infrastructure — not model quality — is now the binding constraint.