Harvey Unveils Tenet, a Legal AI Built on Open-Source Chinese Models

Harvey, a prominent US legal AI startup, has unveiled Harvey Tenet, a specialized legal model built on Moonshot AI’s open-weight Kimi K3 framework using advanced post-training techniques. This milestone marks the first instance of a Chinese open-weight model being integrated into the training pipeline of a top-tier American vertical AI firm.
Previously, Harvey powered its legal services using closed-source models from providers like OpenAI and Anthropic, serving over 2,400 organizations and more than 200,000 lawyers, including numerous global corporate legal departments.
According to public data, Harvey is valued at $1.1 billion with annualized revenue reaching approximately $350 million, establishing it as the highest-valued legal AI unicorn. The company is currently initiating a new funding round at a valuation of roughly $1.55 billion. This strategic shift from closed-source to Chinese open-source foundations underscores increasing practical confidence in open-weight models within specialized markets.
Related article
Andreessen-backed Vals Targets AI Benchmarking Leadership
Benchmarking has become the industry standard for how AI companies validate their models’ capabilities, leverage favorable metrics to differentiate from competitors, and market their superiority. In short, strong benchmark results almost always trans
Apple Exec John Ternus Calls iPhone Best AI Device
Apple’s incoming CEO John Ternus clarified the company’s AI strategy during Wednesday’s Surprise and Shine event, asserting that the iPhone is already the premier AI device and prioritizing user data privacy over competitors.As Ternus succeeds Tim Co
Qwen-Audio-3.0-TTS: Qwen's Large Voice Synthesis Model Cuts First Packet Delay to 300ms
The competition in voice technology among major AI developers has intensified with a new entry. On July 20, Qwen released its latest audio synthesis model, Qwen-Audio-3.0-TTS, shifting control from complex technical configurations to simple natural l
Related Special Topic Recommendations
Comments (0)
0/500

Harvey, a prominent US legal AI startup, has unveiled Harvey Tenet, a specialized legal model built on Moonshot AI’s open-weight Kimi K3 framework using advanced post-training techniques. This milestone marks the first instance of a Chinese open-weight model being integrated into the training pipeline of a top-tier American vertical AI firm.
Previously, Harvey powered its legal services using closed-source models from providers like OpenAI and Anthropic, serving over 2,400 organizations and more than 200,000 lawyers, including numerous global corporate legal departments.
According to public data, Harvey is valued at $1.1 billion with annualized revenue reaching approximately $350 million, establishing it as the highest-valued legal AI unicorn. The company is currently initiating a new funding round at a valuation of roughly $1.55 billion. This strategic shift from closed-source to Chinese open-source foundations underscores increasing practical confidence in open-weight models within specialized markets.
Andreessen-backed Vals Targets AI Benchmarking Leadership
Benchmarking has become the industry standard for how AI companies validate their models’ capabilities, leverage favorable metrics to differentiate from competitors, and market their superiority. In short, strong benchmark results almost always trans
Apple Exec John Ternus Calls iPhone Best AI Device
Apple’s incoming CEO John Ternus clarified the company’s AI strategy during Wednesday’s Surprise and Shine event, asserting that the iPhone is already the premier AI device and prioritizing user data privacy over competitors.As Ternus succeeds Tim Co
Qwen-Audio-3.0-TTS: Qwen's Large Voice Synthesis Model Cuts First Packet Delay to 300ms
The competition in voice technology among major AI developers has intensified with a new entry. On July 20, Qwen released its latest audio synthesis model, Qwen-Audio-3.0-TTS, shifting control from complex technical configurations to simple natural l





Home






