Peter Zhang
Jun 30, 2026 18:56
Claude Sonnet 5, Anthropic’s newest mannequin, is now stay in Harvey, delivering improved authorized activity efficiency for enterprise use.

Anthropic’s Claude Sonnet 5, the most recent iteration of its mid-tier giant language mannequin, is now accessible in Harvey, the AI-driven authorized assistant platform. Formally launched on June 30, 2026, Sonnet 5 goals to ship important enhancements in authorized accuracy and multi-step activity execution, concentrating on each litigation and transactional workflows.
In keeping with Harvey, Sonnet 5 achieved a 5.8% all-pass rating on the platform’s Authorized Agent Benchmark (LAB), a stringent analysis designed to simulate complicated authorized duties at regulation corporations. The mannequin excelled in drafting duties, notably in vitality and pure sources, actual property, and capital markets. On the BigLaw Bench, a separate benchmark targeted on giant regulation agency use circumstances, Sonnet 5 posted a formidable 91.3% rating—surpassing all earlier Sonnet and Opus fashions. Notable strengths included danger evaluation, compliance, and case administration.
“Sonnet 5 brings a significant leap in authorized high quality over Sonnet 4.6,” commented Niko Grupen, Head of Utilized Analysis at Harvey. “In early testing, it was each extra correct and extra exact than its predecessor, delivering stronger solutions in fewer phrases.” Nevertheless, Grupen additionally famous that the mannequin nonetheless faces challenges with dense, specialised analytical duties, reminiscent of tax and structured finance, and stays imperfect in finishing each step of complicated multi-step duties with out error.
Claude Sonnet 5 represents Anthropic’s effort to make superior ‘agentic’ AI capabilities—autonomous activity execution, device utilization, and enterprise workflow integration—extra accessible and reasonably priced. Positioned between the more expensive Mythos- and Fable-class fashions and the older Sonnet 4.5, it balances efficiency with practicality for mainstream enterprise adoption. The mannequin additionally adheres to Anthropic’s Cyber Verification Program, making certain enhanced reliability and security for real-world deployments.
Harvey plans to roll out Sonnet 5 to eligible U.S. prospects within the coming days, with availability within the EU and Australia to observe shortly. As Anthropic continues increasing its AI choices throughout platforms like AWS, Microsoft Foundry, and shortly Google Vertex AI, Sonnet 5 may play a pivotal function in bridging the hole between frontier AI capabilities and on a regular basis enterprise AI deployment.
Picture supply: Shutterstock
