Harvey Trains Custom Model to Cut Costs and Boost Quality in Legal Review
ypatil125 · x · 2026-08-15
Harvey partnered with Applied Compute to train a specialized model for its high-usage Review Table case. The model outperforms frontier models in answer and citation quality while significantly reducing costs, crucial given that a single run can cost $26k.
Related event: Harvey Post-trains GLM-5.2 to Cut Legal AI Costs by 50%(4 posts)→
More from Infra
- RTX 3090 gets 35 t/s on Qwen 3.8 27B — cviperr33 · 2026-08-15
- CME to launch futures contracts tracking Nvidia H100/B100 compute costs — AccBalanced · 2026-08-15
- Feedback: Serverless GPU capacity tight, placement times high — tobowers · 2026-08-15
- Touchmark launches futures marketplace for AI tokens, up to 30% below on-demand rates — ycombinator · 2026-08-15
- Macro Analysis: Is the AI Dip a Buy? Key Levels to Watch — Beth_Kindig · 2026-08-15
- Optimized Dual 3090 Quantization of Qwen3.8-27B Released — luedtek · 2026-08-15