Moonshot's 2.8T-parameter Kimi K3 becomes the largest open-weight model, triggering a chip-stock sell-off
Chinese startup Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model with a one-million-token context window that its internal benchmarks rank behind only Claude Fable 5 and GPT-5.6 Sol. Because the weights are open, any team can run a near-frontier model without a proprietary vendor — a shift that fed a sell-off in AI chip stocks and reopened debate over frontier training costs. Those benchmark numbers are still Moonshot's own and await third-party validation.
Source: news.cgtn.com ↗
Industry benchmarks place the model just behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol.
CGTN
Why this matters
- → Open-weight frontier model removes proprietary vendor gatekeeping on AI capability
- → Chip-stock sell-off signals market uncertainty about frontier training ROI
- → Benchmark claims still unvalidated; real-world performance remains unknown
Open weights reshape AI economics