米醋API

From 4:00 to 10:35 AM (Beijing time) on August 28, a billing anomaly occurred on the platform, causing some requests during this period to be charged higher than normal. The issue was resolved immediately upon detection, and all requests have now returned to normal billing. After calculation, the actual affected charges were only a portion of the consumption during that period. As a gesture of goodwill, instead of refunding the exact difference per transaction, we have uniformly returned 90% of the consumption amount during that period to the account balance of affected users. The refunded amount fully covers the affected portion and exceeds the actual overcharge. The refund has been automatically completed without any application needed. You can log in to the console to view the balance change. We sincerely apologize for the inconvenience caused by this anomaly. We will continue to improve our service quality. Thank you for your understanding and support.

We discovered on the morning of August 28 that a billing anomaly occurred on the platform from 4:00 to 10:30 AM (Beijing time) on August 28, causing some users' requests during this period to be charged higher than normal. After the issue was discovered, we handled it immediately. Billing has now returned to normal, and all subsequent requests are billed at the standard rate. For affected users, we will proactively calculate the overcharged amount during this period and fully refund it to your account balance. No proof or manual application is required. The refund process is expected to be completed soon, and we will notify you separately. We sincerely apologize for the inconvenience caused by this anomaly. Thank you for your understanding and support.

To ensure the relative stability of the group, the vip_2_cheap group will be adjusted to a 0.2x multiplier at 20:00 today.

The grok_heavy group has restored access to the grok-4.6 model.

GPT group price increase notice: Due to recent risk control factors, to maintain service stability and availability, Micu will increase the multiplier for the vip_2 and vip_2_cc groups to 0.5x starting from 2026/08/17 00:00 (next Monday). Once the risk control issues are resolved, we will lower the multiplier as soon as possible. In the future, we may separately open a low-multiplier mixed pool for your choice. If you cannot accept this price adjustment, please submit a refund ticket on the site. We will process refunds with zero fees within about one working day after submission. Thank you for your understanding and support.

Due to the Kiro channel being cut off, the free2 group has been temporarily taken down. Please switch to another group in time. Restoration time TBD.

gpt-5.6-luna has been relisted. This model is currently rate-limited globally. It was relisted only because many review tasks call luna in the background; it is still not supported for use as a primary daily driver model.

Due to business reasons, 米醋 will remove the gpt-5.6-luna model on 2026年08月06日 16:50. After that, the model will no longer be accessible. Restoration time will be announced later. If you need a refund due to this business adjustment, please submit a refund ticket on the site; we will process it within two working days.

The vip_1_max_cheap group has been relisted for testing; the 1x multiplier will be restored once stability is confirmed.

Due to recent issues with the Claude account creation channel, the vip_1_max_cheap group will be temporarily taken down tonight (0722) at 10 PM. Please switch groups in advance. Once the account creation issue is resolved, we look forward to Cheap meeting you all again.

The main site route timeout has been changed to ten minutes.

Due to resource constraints, some models in the discount domestic model group have been temporarily taken down and will be relisted when resources are sufficient. Meanwhile, official API groups for 智谱 and 月之暗面 are now available at 30% off.

The Claude MAX external group has been temporarily restored (the built-in CLI prompt has not been removed; it will be removed after stable testing).

GPT5.6 series models have been pushed.

Claude MAX external related groups are under temporary maintenance. Restoration time will be announced later.

Image group has been relisted.

Image group is temporarily offline for maintenance. Restoration time will be announced later.

The glm-5.2-fast model is now available on the self-deployed channel, billed at 1.5x.

New vip_4_aidc group launched: self-deployed channel backed by the B300 compute center. GLM is available first; other domestic models will be added gradually.