The use case: how Harvey deploys Baseten
Harvey worked with Baseten Research to build a pipeline that post-trains an open-weight 27B model on its Legal Agent Benchmark with a long-horizon legal-agent harness in the loop. Baseten also developed STILL, a technique for compressing the KV cache, as a less lossy alternative to text-based compaction for long legal matters.
Other AI tools used by Harvey
Other public relationships recorded on usedby.
Explore Harvey’s AI stack
Other companies using Baseten
Customers across industries, with public evidence.
EliseAIConfirmed
GammaConfirmed
HebbiaConfirmed
Latent HealthConfirmed
Magicare.aiConfirmed
OpenCodeConfirmedGo straight to the source
Based on Harvey’s own page about how it uses Baseten. We check the public evidence; the results are reported by the company.
harvey.ai- Source published
- May 27, 2026
- Last checked
- Nov 2, 2026
Open source capture What the story claims, and what we checked
What we compared with the page.
- The passage quoted aboveChecked
Copied word for word from the page, near the name of Harvey.
- The numbers in our summaryChecked
Each one is printed on the page.
- The publication dateChecked
Read from the page’s own metadata, never guessed.
- Harvey uses BasetenChecked
Confirmed line.
- The result itselfNot checked
We quote it; we did not measure it.
In collaboration with Baseten Research, we built a pipeline that takes LAB signal, pairs it with a legal-agent harness designed for long-horizon legal work, and post-trains an open-weight model with that harness in the loop.