October 26, 2026 – Kimi has apologized for service issues affecting paid subscribers using the K3.1 Pro model and announced a compensation plan for all active subscription tiers.
In an early-morning statement, the company acknowledged that members had experienced rate limiting and unnecessary quota consumption when running tasks on K3.1 Pro. “The problem is on our side, and we should have fixed it earlier,” Kimi said.
Under the compensation plan, Kimi will reset quotas for all valid subscribers on October 26. Members will receive 25% of their monthly quota as compensation, equivalent to roughly one week of usage. If a member’s current usage is below 25% of the monthly total, their quota will be reset to zero. The updated usage bar is expected to reflect the adjustment by the end of the day.
K3.1 Pro was released on October 24 as Moonshot AI’s most capable model to date, with a 3-trillion-parameter scale and a 1-million-token context window. Shortly after launch, paid users reported that complex agent tasks were exhausting their quotas unusually quickly, sometimes triggering rate limits before tasks were completed. Kimi’s acknowledgment of “unnecessary sub-agent calls” directly addresses those complaints.
On the technical side, Kimi said it has already adjusted its configuration and is reducing unnecessary sub-agent calls and output lengths so that the same quota can complete more actual tasks. The company did not disclose the specific technical cause of the issue or how many users were affected.
The episode highlights a recurring challenge for large-model services: as models gain stronger agent and long-context capabilities, internal call chains and output volumes can increase sharply, making quota design and service stability more difficult to manage.
