Breaking

Moonshot AI Halts New Kimi K3 Subscriptions as Demand Overwhelms Compute Capacity

July 19, 2026 at 16:08 EDT

Foundation Models
Open Source
Software Dev & Coding

Beijing-based Moonshot AI has temporarily paused new subscriptions for its flagship Kimi K3 model, citing an extreme surge in demand just days after the model's July 16, 2026 launch. The company said existing subscribers are unaffected and that it is prioritizing available compute for current members while it brings additional capacity online.

July 16, 2026 · Moonshot AI

Kimi K3 Launches, Then Pauses Sign-Ups as Demand Overwhelms Compute

A 2.8-trillion-parameter open-weight model pushes Chinese AI into frontier territory — but a surge of new users forced Moonshot to freeze new subscriptions days after release, prioritizing existing members while it brings more GPUs online.

2.8T

parameters — first in the ~3T class

token context window

Intelligence Index — top few %

16/896

experts activated per pass

Output pricing: roughly half the cost of rival proprietary models

Cost per million output tokens ($) — shown to scale

$15

Kimi K3

~$30

Typical rival

Input $3 / M · cached input $0.30 / M · output $15 / M

Praised for

Front-end & web engineering — briefly #1 on Frontend Code Arena
Agentic workflows: large repos, tool use, debugging
Cost-effective coding; open-weight gap closing fast

Criticized for

Slow inference — ~26 tokens/sec, long waits
429 errors, quota queues, heavy token use
~2TB VRAM to self-host — impractical for most

What happens next

Demand surge overwhelms GPU capacity

→

New sign-ups paused; existing users unaffected

→

Plans split: Web/App/Work + Kimi Code tier

→

Open weights July 27 — access broadens

A $500M Series C (Jan 2026) funds the compute expansion; sign-ups reopen as more capacity comes online.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.