According to a Financial Times report citing anonymous sources, Chinese AI lab Moonshot AI’s next model in the Kimi series, named Kimi K3, is expected to perform on par with or potentially exceed Anthropic’s Opus 4.8, a closed-source frontier model.
What is reported about Kimi K3
The Financial Times says Kimi K3 will be the largest open-weight AI model from China, with a parameter count reportedly between 2 trillion and 3 trillion. The model is said to be released “in the coming days.”
Moonshot’s previous Kimi K2 models were well received in the open source AI market, scoring highly on benchmarks and demonstrating capabilities that were not far behind the latest top-tier models.
Funding and valuation
The report also states that Moonshot is raising fresh capital in a funding round that would value the company at $31.5 billion. In May, the company raised $2 billion at a $20 billion valuation.
Context: debate over closed vs. open models
The timing of the report coincides with renewed debate over whether companies should pay for expensive closed-source models from providers such as OpenAI and Anthropic. Some industry leaders worry that operators of closed systems could access or extract data that clients submit for use with products like ChatGPT and Claude.
As a result, executives and vendors are pitching alternatives or recommending that companies adopt cheaper open-source models and fine-tune them for internal use. The Financial Times mentions DeepSeek, Z.ai, and Moonshot as examples of providers of such open-source options.
Implications
If the reports are accurate, Kimi K3’s arrival could bring open-weight models closer to the performance of closed-source frontier models and boost interest in open-source solutions. However, independent benchmarks and real-world deployments after the model’s release will be needed to verify its capabilities and practical value.
