Harvey's Gross Margin Went From 50% to Negative 50% on Agent Token Spend, Then Recovered on a Kimi K3 Fine-Tune

Bloomberg·high signalXBlueskyLinkedInCopy link

Bloomberg reports that after Harvey shipped an agent update in March, token consumption rose twentyfold under OpenAI and Anthropic usage-based enterprise pricing, dragging gross margin from about 50% in January to -50% by June. Margins turned positive again in August after Harvey launched an in-house model post-trained on Moonshot's open-weight Kimi K3, though it still routes the hardest tasks to Claude Opus. Abridge is building a clinical foundation model on NVIDIA's open model, and Ramp and Rogo are exploring training their own for the first time.