Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work
Harvey's legal agents just got a capability boost—but only one benchmark held up to scrutiny. Here's what actually changed.
Why it matters
Harvey Tenet represents a post-trained agent model purpose-built for legal work, but the headline performance claims need independent validation. This matters to practitioners evaluating legal agent reliability and to enthusiasts tracking the frontier labs' post-training strategies.
The key facts
11 to knowHarvey Tenet is a Kimi K3 base model post-trained by Harvey with Fireworks AI
Nearly doubles LAB (legal agent benchmark) task completion rate
Only one benchmark number has independent verification as of publication
Purpose-built for long-horizon legal agent workflows
Post-training/specialization strategy for domain-specific agents
Benchmark credibility issue flagged in headline
Harvey Tenet: post-trained model based on Kimi K3, trained via Fireworks AI
Nearly doubled LAB task completion rate (specific metric and baseline not independently verified in article)
Focused on long-horizon legal agent work
Only one benchmark number survives independent verification (article admits gap)
Published on MarkTechPost (secondary source, not Harvey's own announcement)
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Harvey's first post-trained model nearly doubles LAB task completion, but only one benchmark number survives independent verification today