Moonshot AI resets quotas after Kimi K3.1 Pro launch glitches
Beijing-based Moonshot AI apologized for rate-limiting issues following the launch of its Kimi K3.1 Pro model. The company is resetting subscriber monthly quotas by 25% and adjusting parameters to reduce unnecessary subagent calls. This move addresses computing capacity constraints exacerbated by US chip export controls, highlighting ongoing infrastructure challenges for Chinese AI labs competing in the global market.
Key points
- Moonshot AI, a Beijing-based artificial intelligence startup, apologized on Monday for rate-limiting and quota consumption errors during the launch of its Kimi K3.1 Pro model.
- The company is resetting the monthly usage quota for all subscribers by 25%, which it states is equivalent to one week's typical usage, while subscribers with usage below that threshold are reset to zero.
- To mitigate the issue, Moonshot adjusted system parameters to reduce unnecessary subagent calls and limit output length, aiming to optimize token generation efficiency.
- The launch follows previous capacity issues with the K3 model, reflecting broader computing shortages affecting Chinese AI labs due to US chip export controls.
- The adjustment serves as a corrective measure to maintain user satisfaction and competitive viability amidst constrained hardware resources in the Chinese AI sector.
Moonshot AI, a prominent artificial intelligence developer based in Beijing, has issued a public apology for technical glitches encountered during the launch of its latest model, Kimi K3.1 Pro. The company acknowledged that rate limiting and unnecessary quota consumption had disrupted service for users, prompting a corrective action to restore confidence in the platform's reliability.
In a statement posted on its Kimi X account, Moonshot announced it would reset the monthly usage quota for all subscribers by 25%. The company noted that this adjustment is roughly equivalent to one week’s standard usage. Subscribers whose consumption had fallen below the 25% mark were reset to zero to ensure equitable distribution of resources. Additionally, engineering teams adjusted internal parameters to minimize unnecessary subagent calls and reduce output length, thereby lowering the total number of tokens generated per prompt.
These measures address recurring infrastructure challenges, as Moonshot previously faced similar computing capacity constraints following the release of the K3 model. The situation underscores the broader impact of US chip export controls on Chinese AI laboratories, which are currently navigating significant hardware shortages. By optimizing software efficiency and compensating users, Moonshot aims to sustain its competitive position in the global AI market despite these underlying hardware limitations.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.
