Global Markets
Home›Global Markets›Trade & Tariffs›OpenAI-backed legal AI startup builds in-house model o…
OpenAI-backed legal AI startup builds in-house model on Kimi K3
Harvey Tenet is post-trained on Moonshot AI’s open-weight Kimi K3 base to target more accurate complex legal work while aiming to lower inference costs.
A US artificial intelligence startup backed by OpenAI has launched its first in-house model, Harvey Tenet, built on Chinese lab Moonshot AI’s open-weight Kimi K3 foundation, underscoring a broader shift by Western firms toward Chinese open-weight systems as development costs rise, according to SCMP Economy.
San Francisco-based legal tech provider Harvey said its new model was post-trained on top of the Kimi K3 base. Post-training is the process of refining a general-purpose foundation model with specialized datasets to improve performance on specific tasks.
Harvey said the resulting system achieved state-of-the-art performance in complex legal work. The company also framed the move as an example of how open-weight models can be post-trained on industry or corporate data to improve accuracy while potentially reducing inference costs, as noted by AI policy researcher Simon Hedlin, cited by SCMP Economy.