
Chinese AI firm Moonshot released the Kimi K3 model on July 16, an open-weight model with 2.8 trillion parameters and a 1 million-token context window. Hindustan Times reports that Kimi K3 performed…
Chinese AI firm Moonshot released the Kimi K3 model on July 16, an open-weight model with 2.8 trillion parameters and a 1 million-token context window. Hindustan Times reports that Kimi K3 performed on par with or better than some leading closed-weight models, while being priced far more cheaply. The Hindu notes that Moonshot claimed frontier-level performance but acknowledged it still trails Anthropic’s Claude Fable 5 and OpenAI’s GPT 5.6 Sol. The model weights were released on July 27, allowing commercial modifications under a flexible licence.

The Hindu says over 270 companies, including Amazon, Google, Meta, Microsoft, Nvidia, OpenAI and SpaceX, signed an open letter supporting open-weight models to keep innovation in the US. Anthropic did not sign and instead warned of risks from open-weight models, pushing for chip restrictions against China. Hindustan Times reports that chip company executives like Nvidia’s Jensen Huang and Cerebras’s Andrew Feldman have welcomed open-weight models, arguing they will expand rather than reduce demand for computing power.
Anthropic has accused Moonshot of illicit distillation, alleging the Chinese firm used fraudulent accounts to extract capabilities from Claude. The Hindu notes this accusation may boost Moonshot’s reputation by suggesting cheaper access to Claude-grade capabilities. The coming weeks will show whether Kimi K3 marks a lasting shift in AI market dynamics.
Sources (2): hindustantimes.com, thehindu.com
This story was synthesised by AI from the 2 sources linked above.
Updated: this story now draws on 2 sources.