Harvey released Harvey Tenet, a research preview post-trained from the open-weight Kimi K3 base using Fireworks' asynchronous reinforcement learning on long-horizon legal tasks. Harvey reports Tenet completes nearly twice as many tasks on its Legal Agent Benchmark versus base Kimi K3, with gains transferring to Mercor's APEX Agents and Crosby's Redline Bench. Weights, a model card, and an API are not published; Harvey says the work will move "from research to production" in its products.
No score is assigned. Sources and their independence are shown in the citation chain below.