← Back
Harvey AI
Harvey deploys GPT-5.4, scores 91% on legal benchmarks with improvements in drafting and compliance
· releasemodelfeatureperformance · harvey.ai ↗

GPT-5.4 Integration Live

Harvey has made OpenAI's latest frontier model, GPT-5.4, available to eligible customers in the US and EU, with Australian availability to follow. The integration represents a natural extension of Harvey's commitment to bringing cutting-edge AI capabilities to legal practitioners.

Performance Improvements

Early evaluations show meaningful gains across multiple dimensions:

  • BigLaw Bench score: 91.0% vs. GPT-5.2's 89.8%
  • Perfect scores on risk assessment and compliance tasks
  • Strong performance in deal management, drafting, and litigation filing analysis
  • Improved drafting format and organizational structure compared to prior versions

Key Strengths

Harvey's Applied Legal Research team identified several practical advantages for legal workflows:

  • Substantive accuracy: The model delivers stronger accuracy in legal analysis and reasoning
  • Drafting precision: Direct engagement with tasks without unnecessary preamble or hedging language
  • Clarity and organization: Intuitive structure with appropriate levels of detail for legal documents
  • Speed: Fast processing suitable for high-volume legal work

Deployment Timeline

GPT-5.4 will be rolled out through Harvey's model selector to eligible clients in the coming days. The model is expected to excel in drafting-intensive work and transactional matters where precision and attention to contractual detail are essential. Customers in the US and EU can access it immediately upon rollout; Australian availability will be announced as regional capacity permits.