← Back

Wire · opportunities

MoonshotAI Ships Kimi K3, First Open 3T-Class MoE Model

Published

28 July 2026

Topic

opportunities

Sectors

AI Frontier Models

Geography

China

Source

Read at aiweekly.co

Verified

Fusion42 · 28 July 2026 · Fusion42 review

Moonshot released Kimi K3, an open-weights 2.8 trillion parameter mixture-of-experts model, claiming 2.5x scaling efficiency over its predecessor through architectural innovations including Kimi Delta Attention and Stable LatentMoE. The release includes open-source infrastructure (attention kernels, MoE communication library, agent environment tools) alongside model weights, potentially lowering deployment barriers for trillion-scale MoE systems.

This Wire brief sits within Fusion42's coverage of AI Frontier Models. Wire is Fusion42's founder-focused intelligence feed: each story is connected to the funds and startups it names — every one with a live profile on Raise or Scout — so founders can follow the capital and the momentum behind the headline rather than just the headline itself. Wire analysis is one of the live surfaces Arthur reasons over.

◆ The Wire takeaway

Open infrastructure for trillion-scale models just became portable. If you're building on top of frontier LLMs, Moonshot has shipped the plumbing—attention kernels and MoE communication stacks—that lets you run K3 outside their datacenters; that's permission to stop waiting for OpenAI's API and ship your own stack.

Related on Wire

Topics

AI Frontier Modelsopen-weights-llmmixture-of-expertsscaling-efficiencyinference-infrastructurechina-ai