
Moonshot AI releases Kimi K3 open weights and infrastructure after shaking up the frontier model race
Quick Answer
Moonshot AI has released Kimi K3's open weights and infrastructure, achieving 2.5 times more intelligence per compute unit.
Quick Take
While it competes closely with models like Fable 5 and GPT-5.6 Sol on benchmarks, independent tests reveal significant gaps in cyber capabilities and math skills, suggesting reliance on distillation techniques.
Key Points
- Kimi K3's model weights and infrastructure are available on Hugging Face.
- The model scores closely to Fable 5 and GPT-5.6 Sol at lower costs.
- Independent tests highlight significant gaps in Kimi K3's cyber and math capabilities.
- Moonshot AI's architecture claims improved intelligence per compute unit.
- The technical report for Kimi K3 is accessible on GitHub.
📖 Reader Mode
~1 min readChinese 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.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.

