Chinese lab Moonshot AI just released Kimi K3, an open-weights model that sets new highs for both Chinese and open-source models and is competitive with frontier systems like Claude Fable 5 and GPT-5.6 Sol at far lower prices.
Kimi K3 features a 1M context window and beats both Fable and Sol on benchmarks for web research, spreadsheet work, frontend design, and long coding. The model also lands at a 57 on the AA intelligence index, sitting behind just Fable (60) and Sol (59) and jumping double-digit spots from the previous K2.6 release.
In one demo, K3 worked solo for 48 hours to design and verify a tiny chip that runs a mini version of itself, hitting 8,700 tokens a second in simulation. That is a striking display of what an open-source model can do when given real engineering tasks and enough compute time.
Pricing also makes K3 accessible. It matches Claude 5 Sonnet at $3 per million input tokens and $15 per million output tokens. Moonshot says the model’s weights will be published by July 27, meaning any developer or researcher can fine-tune or build on it without paying for a closed API.
Why this matters
Pull right up on Fable and Sol’s frontier tail is no small feat, especially for a soon-to-be open-weights model at this price point. Dario Amodei’s “6-12 months behind” estimation for both China and open source suddenly looks more like just a single release cycle.
The frontier crown used to be reserved for Anthropic and OpenAI. That table now has a third seat. For authors and creators watching the AI space, the real signal is speed. What costs dollars today on a closed frontier model will have a free, open alternative within weeks. Builders who ignore that shift will spend the next quarter retooling around models that were state-of-the-art for only a few months.
Bottom line: if your book content, marketing, or research workflow leans on AI tools, start testing Kimi K3 the moment weights drop on July 27. The performance is already close enough to make the switch worth the effort, and the cost savings are significant at scale.