Legal AI Leader Harvey Expands Beyond OpenAI with Anthropic and Google Models

Harvey, a prominent legal AI startup initially supported by the OpenAI Startup Fund, has announced a significant expansion in its AI model usage. Moving beyond exclusive reliance on OpenAI’s foundation models, Harvey will now integrate leading models from Anthropic and Google. This strategic shift reflects Harvey’s commitment to leveraging the best AI technologies available to enhance its legal AI solutions.

Harvey’s decision follows extensive internal benchmarking using its proprietary BigLaw benchmark, which evaluates AI models on a variety of legal tasks. The results demonstrated that different foundation models excel in different areas—for example, Google’s Gemini 2.5 Pro is highly effective at legal drafting but less adept at complex pre-trial tasks, whereas OpenAI’s models perform better in those pre-trial scenarios. Anthropic’s Claude 3.7 Sonnet also showed competitive performance.

By embracing a multi-model approach, Harvey aims to fine-tune these diverse foundation models specifically for the legal market, optimizing performance across a broad spectrum of legal applications. This approach also supports the development of AI agents capable of handling complex legal reasoning tasks more effectively by leveraging the unique strengths of each model.

Harvey’s move is particularly notable given its history as one of the first startups backed by the OpenAI Startup Fund and its valuation exceeding $3 billion following recent funding rounds led by Sequoia and Google’s GV. Despite Google’s early investment, Harvey had not previously adopted Google’s AI models, making this integration a significant milestone that highlights the competitive and collaborative dynamics in the AI landscape.

In addition to adopting multiple foundation models, Harvey plans to contribute to the AI community by publishing a public leaderboard showcasing model performance on legal tasks. This leaderboard will feature nuanced insights from top legal professionals, providing a richer understanding of AI capabilities beyond simple benchmark scores. This transparency promotes accountability and drives innovation across the legal AI sector.

Harvey’s CEO Winston Weinberg emphasized the company’s ongoing collaboration with OpenAI while expressing enthusiasm for expanding customer options through diverse AI models. This balanced approach underscores the evolving nature of AI partnerships and the importance of leveraging multiple technologies to meet complex industry needs.

This development highlights broader trends in AI adoption where specialization and multi-vendor strategies are becoming key to delivering superior outcomes, especially in specialized fields like legal technology. Harvey’s approach exemplifies how startups can harness competitive AI ecosystems to drive innovation and better serve their markets.