2026-10-11 17:12 UTC

Kimi will continue rationing K3 access or restricting new subscriptions until added inference capacity catches up with reported demand.

state: expiredheat: lowuncertainty: mediumconvergesscott: mediumkimi-k3 inference-capacity model-demandKimiMoonshot AI

What is this?

Kimi, Moonshot AI’s AI service, temporarily paused new K3 subscriptions after demand over a 48-hour period pushed its GPU capacity close to its limits. The company said it is prioritizing compute for existing subscribers, who remain unaffected, while adding capacity and planning to reopen subscriptions in batches. The supplied evidence supports a temporary restriction, but does not establish how long rationing will continue or when added capacity will arrive.

Why it matters to Scott

Kimi’s temporary capacity rationing reinforces Scott’s existing use of provider-neutral gateways, explicit fallback paths, and cost-tiered routing as protection against constrained model access. However, no hit shows that his systems depend on Kimi/K3, and the evidence does not establish sustained API scarcity, so this is only another example of a pattern he already builds around rather than an actionable change.
dev:concept.task-aware-model-routingdev:concept.cost-tiered-llm-routingdev:technology.litellm
queries asked of Scott's wikis
  • inference capacity bottlenecks and demand rationing
  • provider throttling and coding-agent reliability
  • multi-provider failover for agent harnesses
  • hosted versus local inference economics
  • AI subscription scarcity and tiered compute access
  • Jevons paradox in AI inference demand

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (9) — ⭐ canonical anchor

sourceobjectauthorscorecomments
🟠 reddit ⭐Kimi is temporarily pausing new subscriptions and prioritizing compute for current members due to surging demand.
singularity
SuggestionMission5162051227
🟠 redditKimi K3 released on web and app
LocalLLaMA
External_Mood4719637261
🟠 redditAnthropic dropped the paywall, say thank you, Kimi!
artificial
StudentSweet360121
🟠 redditFor those who didn't get to join Kimi, there is a waitlist now.
singularity
Fabulous_Bonus_8981268
🟠 redditOllama Kimi-K3 cloud model use “not included in plan usage” can only use via separate pay-per-token “extra usage” API
LocalLLaMA
Porespellar315
🟧 hnKimi K3 Now Available via Telnyx Inference APIfionaattelnyx13187
🟠 redditWhat's the cheapest way to get Kimi K3? (NOT necessarily the fastest)
LocalLLaMA
Tank_Gloomy067
🟧 hnMoonshot built on 20k Nvidia chip cluster from Alibabagk111376
🟠 redditKimi K3 full model running on 16x GB10 cluster at 20+tps
LocalLLaMA
ciprianveg1865353

Interpretation history

Decision trace