Harvey

Harvey is the legal-AI company used in 171: 【AI季报 26Q2】从 coding 到 RSI，强者愈强的未来？ as the concrete case for Enterprise Owned Models. The source says Harvey worked with Applied Compute to post-train a legal-domain model based on GLM 5.2 or a related GLM line, and that the resulting model performed strongly on Harvey’s legal-agent benchmark.

The episode’s point is strategic rather than purely legal: companies with high-value proprietary data, clear evaluation systems, and frequent expert workflows may prefer owning or post-training a domain model instead of relying only on OpenAI or Anthropic frontier services.

Connections

• Applied Compute — post-training partner in the source.
• GLM 5.2 and Zhipu AI — open-model ecosystem connected to the case.
• Enterprise Owned Models — concept this case motivates.
• Legal AI Hallucination and Human-In-The-Loop Legal AI — legal-AI quality and responsibility context.