😼 Kimi K3 goes open
The Neuron · Grant Harvey · 2026-07-17
Moonshot AI releases Kimi K3, a 2.8-trillion-parameter open-weight model with a one-million-token context window that benchmarks near frontier models including GPT-5, with full weights due publicly by July 27.
Appears in
Extraction
Topics: open-weights-modelschinese-aifrontier-modelsmodel-benchmarksopen-source-ai
Claims
- Moonshot AI released Kimi K3 with 2.8 trillion parameters, native vision, and a 1M-token context window, with open weights due by July 27.
- Kimi K3 reportedly beat GPT-5.6 Sol on BrowseComp and Automation Bench benchmarks according to Moonshot's own testing, and Artificial Analysis placed it one point ahead of Claude Opus 4.8 on its intelligence index.
- Kimi K3 uses sparse mixture-of-experts architecture, activating only subsets of parameters per task to reduce inference cost despite its enormous total parameter count.
- Open Chinese AI models benefit the global ecosystem because US and other companies can adopt and customize them as a private base layer, making government restrictions counterproductive.
- Open-weight frontier models erode closed US labs' pricing and distribution leverage, benefiting companies that rely on AI infrastructure.
Key quotes
if a Chinese open model comes out that is near frontier quality, this is good for everyone, because global companies (including US ones!) get another powerful base layer they can adapt and run privately.
when a new frontier level open model comes out, U.S. labs lose some pricing and distribution leverage. This is also good for American companies whose engineering departments are like frogs about to be boiled alive when these companies go public.
Closed labs selling polished assistants versus open ecosystems selling control, customization, and compounding advantage. Which system do you want to build your company on top of?