Breaking

Moonshot AI's Kimi K3 Tops Anthropic's Opus 4.8 on Key Benchmarks, Narrowing US-China AI Gap

July 17, 2026 at 05:55 EDT

Foundation Models
Open Source
Policy & Regulation

China's Moonshot AI has released Kimi K3, a 2.8-trillion-parameter model that outperforms Anthropic's Opus 4.8 on several benchmarks and undercuts it on price by 40%, in the latest sign that America's lead in frontier AI has all but evaporated. The model launched around July 16, 2026, and is scheduled to ship an open-weight version on July 27 under a Modified MIT license, according to Axios.

July 16, 2026 · Moonshot AI

Kimi K3 Erases America's Frontier-AI Lead

China's Moonshot AI has released a 2.8-trillion-parameter model that beats Anthropic's Opus 4.8 on several benchmarks — and undercuts it on price by 40%. An open-weight version ships July 27.

2.8T

parameters in Kimi K3

token context window

on Artificial Analysis (score 57)

Price to run vs. Anthropic Opus 4.8

100%

Opus 4.8

60%

Kimi K3

Near-Opus performance at a 40% discount.

Benchmark standing

Edged out Anthropic's Fable 5 & OpenAI's GPT-5.6 Sol on a front-end coding test (AI Arena)
Surpassed Anthropic's Opus 4.8 in text rankings
Native visual understanding, built for software engineering & deep reasoning

SUPPORTERS SAY

Strong leaderboard placement and near-Opus quality at a substantial discount — a standout release, soon with open weights.

SKEPTICS SAY

Sluggish inference, cautious and clumsy agentic behavior, repetitive loops — plus distillation allegations from Anthropic over Opus-like output.

The shift

Open weights + a long context window + aggressive pricing could accelerate adoption and pressure US labs still relying on closed, API-only frontier models. Full technical report expected around the July 27 open-weight release.

Continue reading

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