Kimi K3: The 2.8 Trillion Parameter Open-Weight Model That Moved Markets
Summary
On July 17, TSMC fell 7% despite a 77% profit jump, SoftBank dropped 9%, and Nvidia briefly lost its most-valuable-company crown. The trigger: Moonshot AI's Kimi K3, a 2.8T-parameter model with a 1M-token context window, second overall on Artificial Analysis behind only Claude Fable 5, first on blind frontend-code testing, at $15 per million output tokens versus $50 for the frontier. Weights go public July 27. After DeepSeek V4 in April, the assumed six-month US-China gap is now measured in weeks.
About this video
On July 17, TSMC fell 7% despite a 77% profit jump, SoftBank dropped 9%, and Nvidia briefly lost its most-valuable-company crown. The trigger: Moonshot AI's Kimi K3, a 2.8T-parameter model with a 1M-token context window, second overall on Artificial Analysis behind only Claude Fable 5, first on blind frontend-code testing, at $15 per million output tokens versus $50 for the frontier. Weights go public July 27. After DeepSeek V4 in April, the assumed six-month US-China gap is now measured in weeks.