← Back

Wire · market

China's Kimi K3 Is Out—And Beats Claude Fable and GPT 5.6 Sol on Key Benchmarks

Published

18 July 2026

Topic

market

Sectors

AI Frontier Models

Geography

China

Source

Read at decrypt.co

Verified

Fusion42 · 18 July 2026 · Fusion42 review

Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model that beats Claude Fable 5 on writing and code generation benchmarks whilst priced identically to Claude Sonnet 5. Full model weights release by 27 July under modified MIT licence, making it the largest freely available AI model in history.

This Wire brief sits within Fusion42's coverage of AI Frontier Models, and 34 sources have reported it between 18 Jul 2026 and 7 Aug 2026.

◆ The Wire takeaway

If you're building on Claude or GPT APIs for writing or code work, you now have a free alternative that scores higher on the same tasks and costs nothing to run locally after 27 July. The economics of your AI stack just inverted.

Coverage

34 sources · first reported 18 Jul 2026 · latest 7 Aug 2026

Related on Wire

Topics

AI Frontier Modelsopen-source-modelsllm-benchmarksai-pricingchina-aimodel-weightscoding-capability