Harvey, a prominent legal AI tool, has announced a significant expansion of its model offerings, now incorporating sophisticated foundation models from both Anthropic and Google alongside its traditional reliance on OpenAI, as detailed in a recent blog post. This shift marks a noteworthy development for the startup, which was one of the earliest beneficiaries of the OpenAI Startup Fund, intended to support companies leveraging OpenAI’s technology. While Harvey reassures that it remains committed to OpenAI, embracing additional models from competitors signifies a notable strategic pivot in the competitive AI landscape.
Founded with early support from the OpenAI Startup Fund in December 2022, Harvey’s trajectory has been impressive, achieving a valuation of $3 billion following a $300 million Series D funding round in February led by Sequoia and backed by other prominent investors. Notably, Google’s investment arm, GV, played a role in Harvey’s $100 million Series C financing, although the startup did not initially adopt Google’s AI solutions until now.
Harvey’s decision to diversify its model offerings stems from its internal benchmarking system called BigLaw, which demonstrated that various foundation models are becoming increasingly proficient at legal tasks, each excelling in different areas. Instead of investing heavily in training its own models, Harvey has recognised the potential to leverage high-performing models from external sources, fine-tuning them for legal applications. This strategy will also enhance its capability to develop AI agents.
Recent testing indicated that several non-OpenAI models outperform Harvey’s previously benchmarked systems, showcasing the rapid advancement in AI capabilities. For instance, Google’s Gemini 2.5 Pro was found to excel in legal drafting but struggled with pre-trial tasks due to its limitations in understanding complex evidentiary rules. Conversely, OpenAI’s model performed well in those pre-trial tasks, illustrating the varied strengths of different models.
In a bid to contribute to the broader AI community, Harvey plans to publish a public leaderboard assessing the performance of various reasoning models in legal contexts. This leaderboard will go beyond simplistic scoring systems, offering nuanced insights from top legal professionals that reflect model performances more thoroughly.
This shift not only highlights Harvey’s commitment to harnessing the best available technology but also applies pressure on its existing backers, including OpenAI and Google, to demonstrate their ongoing value in this rapidly evolving sector. Harvey’s CEO, Winston Weinberg, expressed gratitude for OpenAI’s support while emphasising the importance of providing diverse options to serve their global customer base more effectively.
Fanpage:Â TechArena.au
Watch more about AI – Artificial Intelligence
