← Back

Moonshot AI

companyCredibility: 57%

Why this score? Chinese AI lab (Kimi models); primary for its own releases, motivated on capability claims.

Tracked Statements (2)

It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.✓

— Moonshot AI · Jul 27, 2026
Context: Core holds: the published weights are for a 2.8-trillion-parameter model, larger than any prior open-weight release, corroborated by Tom's Hardware and, pre-release, by ANU's Rahul Shome in Nature. Caveats that don't falsify it: '3T-class' rounds up 2.8T, and only 104B parameters activate per token. Capability is independently corroborated — Artificial Analysis scores K3 at 60 — but Moonshot's own benchmark table, run under its Kimi Code harness, is not.

Moonshot AI claims Kimi K3 beat Claude Opus 4.8 and GPT 5.5 — models CNBC describes as sitting just behind Anthropic's and OpenAI's leading-edge systems — on benchmarks including coding and general agents.±

— Moonshot AI · Jul 17, 2026
Context: On the re-based Artificial Analysis index K3 (60) now leads both models it named — GPT-5.5 (56) and Claude Opus 4.8 (57). Mixed overall because the specific coding and general-agent wins still rest on Moonshot's own benchmark table, run under its Kimi Code harness while rivals use theirs, and are not independently replicated; and neither model named was its lab's flagship — Claude Opus 5 now tops the index at 63.