OpenAI-backed legal tech firm Harvey pivots to Chinese Kimi K3 open-weight model
On Thursday, Harvey, a legal tech company located in San Francisco and supported by OpenAI, Sequoia Capital, and Andreessen Horowitz, revealed its latest model, Harvey Tenet, which has been post-trained using the open-weight Kimi K3 base. The firm asserts that this new system delivers 'state-of-the-art' outcomes in intricate legal tasks. This development underscores a trend among Western tech companies that are leveraging Chinese open-weight models to cut down on development expenses and enhance precision. AI policy researcher Simon Hedlin remarked on social media that Harvey's initiative exemplifies how open-weight models allow developers to fine-tune systems with industry-specific data. This post-training process involves enhancing a general model with specialized datasets to perform particular functions more effectively.
Key facts
- Harvey, a San Francisco-based legal tech provider, announced its new model Harvey Tenet on Thursday.
- Harvey Tenet was post-trained on top of the open-weight Kimi K3 base.
- Harvey's backers include OpenAI, Sequoia Capital, and Andreessen Horowitz.
- Harvey claims the system achieved 'state-of-the-art' performance in complex legal work.
- AI policy researcher Simon Hedlin praised the move as an example of open-weight models enabling post-training for specific data.
- Post-training is the process of refining a general-purpose base model with specialized data sets.
- There is a growing shift by Western tech firms towards Chinese open-weight systems amid soaring development costs.
Entities
Institutions
- Harvey
- OpenAI
- Sequoia Capital
- Andreessen Horowitz
- Kimi K3
Locations
- San Francisco
- United States
- China