Kimi resets 25% of monthly quota for all paid tiers after K3.1 Pro rate-limit and quota problems

Kimi has apologized to paying members for rate limiting and unnecessary quota consumption on K3.1 Pro, and will give every active subscriber 25% of their total monthly quota back. The statement was posted in the early hours of October 26, according to IT Home and National Business Daily, as reported by Pandaily.

"We apologize for the rate limiting and unnecessary quota consumption that Kimi members have recently experienced when using K3.1 Pro," the statement said. "The problem was on our side, and we should have fixed it sooner."

What Kimi is changing

• One-time quota reset: applies to subscription members on all tiers. Every active member receives 25% of their total monthly quota, which Kimi describes as roughly one week of usage.
• Low-usage members: anyone who has used less than 25% of their monthly quota will have usage reset to zero.
• Timing: Kimi said the adjusted numbers would appear on users' usage bars before the end of the day.
• Configuration change: Kimi says it is reducing unnecessary sub-agent calls and output length, so the same quota can complete more real tasks.

In agent-style workflows, a single request can spawn several sub-agents and long outputs, and each draws on a member's allowance. The configuration change targets that consumption pattern.

Background: K3.1 Pro launch

Moonshot AI launched K3.1 Pro on October 24, the first model in its K3.1 series. The company calls it its most capable model so far, with about 3 trillion parameters and a 1 million token context window. A Highspeed version is also available, which Moonshot AI says produces output about two to three times faster.

The models are available on kimi.com, the Kimi app, Kimi Work, Kimi Code and the Kimi API. Moonshot AI says full model weights will be released before November 4.

The company reported a score of 65.7% on Terminal-Bench 4.0, a test of coding agents in real terminal environments. That figure is self-reported.

Earlier capacity strain

In July, days after the K3 launch, Kimi temporarily paused new consumer subscriptions as demand outran available compute.

What this means

Kimi attributes the problem to its own side and has paired the quota reset with a configuration change, not only compensation. The sub-agent and output-length reductions suggest that quota burn on K3.1 Pro was driven partly by how much work each agent request generated. For members using Kimi Work or Kimi Code, the same quota should now stretch across more completed tasks, though Kimi has not published figures on how much. The episode also follows the July subscription pause, so capacity during new model launches remains a recurring issue for Kimi.