Harvey's Tenet: a post-trained Kimi K3 that finishes 2x more legal tasks
Harvey just released Tenet, its first post-trained open-weight model: a Kimi K3 base tuned with asynchronous RL inside realistic long-horizon legal environments. On LAB it completes almost twice as many held-out tasks as the raw base and tops LAB Contracts, and gains transfer to unseen benchmarks like Redline Bench. Separate tracks cover diligence, review tables, and firm knowledge. Reward shaping also pushes toward cheaper inference: better legal agents at lower per-token cost.