Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work
Harvey launched Harvey Tenet, a legal agent model post-trained on Kimi K3 via Fireworks. The company claims nearly doubled LAB task completion, though independent verification remains limited.
Harvey has released Harvey Tenet, a specialized large language model designed for long-horizon legal tasks. The model utilizes the Kimi K3 base architecture and underwent post-training facilitated by Fireworks.
According to the release, this iteration significantly improves performance on legal agent benchmarks. Harvey reports that task completion rates on the LAB benchmark nearly double compared to previous versions, highlighting advancements in reasoning over extended workflows.
However, independent validation of these performance metrics remains sparse. Currently, only one benchmark number has survived external verification, suggesting caution is warranted when evaluating the claimed improvements against industry standards.
This development underscores the trend of specialized post-training for vertical AI applications. By focusing on legal workflows, Harvey aims to enhance reliability in complex, multi-step agent tasks within the legal technology sector.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.