Sonnet 4.5, Now Live in :Harvey:
Announcing Sonnet 4.5 in Harvey
Sep 30, 2025
Harvey Team
Today we are making Claude Sonnet 4.5 available in Harvey.
Claude 4.5 Sonnet represents the latest frontier reasoning model from Anthropic, which was trained with an emphasis on code generation and numerical analysis. Early evaluations point to these aptitudes translating to meaningful gains in autonomous task completion, research, and numerical reasoning. We see these gains translate directly into legal workflows where those skills are critical, such as deal management and due diligence.
We’ve also observed improvements in long-context reasoning, which makes Sonnet 4.5 particularly well-suited for analyzing complex legal documents like Credit Agreements. On our BigLaw Bench evaluation suite, Sonnet 4.5 achieved the highest score yet for the Claude family of models at 89.55%, further underscoring its potential to handle the rigor and nuance of elite legal practice.
“Our goal is always to ensure that Harvey customers have the best models for each discrete element of their work in professional services. By launching Sonnet 4.5 in Harvey today, we are expanding that commitment and ensuring that our customers globally have access to the models that work best for their unique business needs.”
Niko Grupen
Head of Applied Research
We’ll be rolling out Sonnet 4.5 in the model selector to eligible clients in the US and EU.