Moonshot AI’s Kimi apologized for rate limits and unnecessary quota consumption on K3.1 Pro, saying the problems were its responsibility and should have been fixed sooner. The company announced a one-time quota adjustment for active paid subscribers across all tiers, alongside configuration changes intended to reduce avoidable usage.
3
12
How the 25% quota adjustment works
Kimi said the adjustment would restore usage equal to 25% of a subscriber’s monthly allowance, which the company described as roughly one week of use. It applied across subscription tiers to members whose subscriptions were still active.
1
3
The reset depended on how much quota a member had already used:
- If usage was at least 25%: the adjustment restored 25% of the monthly allowance.
- If usage was below 25%: the used-quota counter would reset to zero.
That second case was not a full monthly reset for every subscriber; it cleared the counter for members who had used less than the 25% threshold.
1
3
Kimi said its usage bar was expected to reflect the change by the end of the day Pacific time. That was the announced deadline, not confirmation that the update was completed on schedule.
3
12
Why Kimi said it was changing configuration
Kimi also said it had adjusted the plan configuration to reduce unnecessary sub-agent calls and output length, with the aim of making the allowance go further.
12
In an agent workflow, a request can involve multiple model calls and generated text—not just the final response. Calls made by sub-agents and their output can therefore add to usage. Cutting unnecessary calls and overly long output is intended to reduce that overhead; it does not establish how much quota users will save in practice. Kimi’s announcement described the changes, but the available reporting does not verify their eventual effect.
12
The K3.1 Pro launch context
The apology followed K3.1 Pro’s rollout. Kimi’s documentation describes it as the first model in the K3.1 series, a 3-trillion-parameter-class flagship with a 1-million-token context window.
17
18 Moonshot also advertised a Highspeed version as delivering roughly two to three times the speed, and listed access through Kimi.com, its mobile apps, Kimi Work, Kimi Code and the Kimi API.
18
30
Moonshot said the full model weights would be released by November 4. That was a stated release commitment, not evidence that the weights were already available when Kimi announced the quota adjustment.
30
The immediate response was therefore two-part: compensate active paid members for a portion of their monthly quota and change configuration to curb usage Kimi described as unnecessary. The company acknowledged the service problem, but the promised usage-bar update and the practical impact of the configuration changes should be treated as announced plans rather than independently confirmed outcomes.
3
12