China's Moonshot AI unveils Kimi K3, a 2.8 trillion parameter model it says rivals OpenAI and Anthropic

1:38 AM · July 17, 2026

China's Moonshot AI unveiled Kimi K3, a 2.8 trillion parameter model the company says rivals leading systems from OpenAI and Anthropic, according to CNBC, making it among the largest open weight AI models released anywhere to date. On the GDPval-AA v2 benchmark, Kimi K3 scored 1,687, placing it third overall behind Claude Fable 5 Max and GPT-5.6 Sol Max, but ahead of Claude Opus 4.8, and on a separate agentic benchmark it beat GPT-5.6 Sol Max while trailing only Fable 5 Max. In real world task automation tests, Kimi K3 ranked first in four out of eight benchmarks tried, and Arena separately ranked it first in a frontend coding evaluation, ahead of Fable 5. The model carries a 1 million token context window and is priced aggressively, at 0.30 dollars per million cache hit input tokens, 3 dollars per million on cache misses, and 15 dollars per million output tokens, well below what comparable closed models typically charge. Moonshot describes K3 as open, though the full model weights are not scheduled for release until July 27. The launch continues a pattern in which Chinese open weight labs have closed the gap with the best funded American frontier labs within months rather than years, and comes as investors and rivals alike are increasingly treating Chinese open models as legitimate competitors rather than a step behind.