Chinese AI company Moonshot AI has released the model weights and technical report for Kimi K3. Along with the model weights on Hugging Face, the company is open-sourcing parts of its infrastructure, including high-performance attention kernels, an MoE communication library, and tools for running AI agents at scale. Moonshot AI claims the new architecture delivers 2.5 times more intelligence per unit of compute. The technical report is available on GitHub.

Since its initial announcement in mid-July 2026, Kimi K3 has caused a stir by scoring close to Western frontier models such as Fable 5 and GPT-5.6 Sol on popular benchmarks, but at a slightly lower cost and now with open weights. However, an independent test by the UK's Cyber Institute found that the model's cyber capabilities lag far behind those of frontier models. The same is true of its math skills.

Both gaps could suggest that Kimi K3 relies on distillation, a technique in which a smaller model learns from the outputs of a more capable one. Chinese models often face this accusation. At the same time, American open-weight advocates increasingly view distillation as a legitimate technique.

AI News Without the Hype – Curated by Humans

Kimi