Chinese lab Moonshot AI released Kimi K3 on July 16 - a roughly 2.8-trillion-parameter open-weight mixture-of-experts model with a 1-million-token context window and always-on reasoning - which it and early benchmarks call the largest open-weight model ever and competitive with the best closed US systems. K3 took first place on the Frontend Code Arena coding leaderboard, passing Anthropic's Claude Fable 5, OpenAI's GPT-5.6 Sol and Zhipu's GLM-5.2, and ranked third on the GDPval-AA real-world-task benchmark, at an API price of $3/$15 per million input/output tokens - a fraction of its US rivals. Moonshot said it will publish the full model weights on Hugging Face under a modified-MIT license around July 27.
/ Sources
/ About this story
Compiled by Venture Atlas from the sources above, using automated AI-assisted research. This is a summary of reporting published elsewhere, not original reporting - follow the source links for the full account. See our editorial standards.
Something wrong here? Email apostructura@gmail.com and we will fix it.
/ Related
- OpenAI says it disrupted a Moonshot-linked distillation campaignWednesday, September 30, 2026
- Moonshot connects Kimi to Wall Street data feedsSunday, September 20, 2026
- CISA warns of industrial-scale China AI distillationWednesday, September 9, 2026
- Moonshot AI files confidentially for $3 billion Hong Kong IPOThursday, September 3, 2026
