China's Moonshot AI apologized on October 25, 2026 for rate limits and unnecessary quota consumption that hit subscribers after the Kimi K3.1 Pro launch — and offered a make-good: a one-time reset restoring 25% of every subscriber's monthly allowance, which the company said equals roughly one week's usage. Subscribers who had used less than 25% would see their recorded usage reset to zero.

The apology, posted on Kimi's X account and reported by Reuters on October 26, came just two days after K3.1 Pro launched across Kimi's web, work and coding products and its API. Moonshot said it had also adjusted subscription settings to cut unnecessary sub-agent calls — helper AI runs delegated part of a larger job — and to shorten output length, two ways a single task could quietly burn through a monthly allowance.

The wording was unusually direct. "These were our issues, and we should have fixed them sooner," Kimi said in its apology, according to the reports. Moonshot had previously run into computing-capacity problems after the earlier K3 launch, and Chinese AI labs broadly face compute shortages following US chip export controls — so a launch-week squeeze was not entirely unexpected, but subscribers clearly felt it anyway.

For paying users, the episode is a reminder that a model's launch-day benchmarks are only half the story: rate limits and quota mechanics decide how much of the model you actually get to use. Whether the trimmed sub-agent calls and shorter outputs make allowances last noticeably longer in practice is still unconfirmed — Moonshot has not published before-and-after quota figures — but the 25% reset is concrete, and it lands before the full model weights are due on November 4.

Sources