Popular legal AI tool Harvey will now be using leading foundation models from Anthropic and Google, moving beyond strictly using OpenAI’s, Harvey announced in a blog post on Tuesday.
This is noteworthy because Harvey is one of the OpenAI Startup Fund’s most successful early-backed portfolio companies. The OpenAI Startup Fund is an OpenAI-associated fund to back companies developing products on top of AI technologies, primarily OpenAI’s own. While Harvey says it’s not abandoning OpenAI, merely adding more models and clouds, this is still a huge coup for OpenAI’s big competitors.
Harvey is one of the first four startups that the OpenAI Startup Fund backed, it said in December 2022. That was back when OpenAI CEO Sam Altman was still running the fund. (The others in that first cohort include Descript, Mem, and Speak.)
Harvey, which has grown like mad since then, is now a $3-billion-valuation startup, it said in February, when it announced a $300 million Series D led by Sequoia, with other big names like Coatue, Kleiner Perkins, and the OpenAI Fund piling in.
Interestingly, Google’s venture arm, GV, led Harvey’s $100 million Series C in July 2024 (and the OpenAI Fund participated in that round, too). But Harvey didn’t immediately adopt Google’s AI models after it put Google’s corporate venture firm on its cap table. (GV participated in Harvey’s Series D as well.)
So, what convinced Harvey to move beyond OpenAI’s models now? The startup’s internally developed benchmark, dubbed BigLaw, showed that a wide variety of foundation models are growing increasingly adept at a range of legal tasks and some are better at specific tasks than others.
Instead of spending its efforts training models, Harvey figured, it could simply embrace high-performing, reasoning foundation models from other vendors (e.g. Google and Anthropic via Amazon’s cloud) and then fine-tune them for the legal market.
Using a variety of models will also help as Harvey creates AI agents, the company says.
“In less than a year, seven models (including three non-OAI models) now outperform the originally benchmarked Harvey system on BigLaw Bench,” Harvey wrote in the blog post.
Harvey’s benchmark also showed that different foundation models are better at specific legal tasks than others. For instance, it says Google’s Gemini 2.5 Pro “excels” at legal drafting but “struggles” with pre-trial tasks like writing oral arguments because the model doesn’t fully understand “complex evidentiary rules like hearsay.”
OpenAI’s o3 does such pre-trial tasks well, according to Harvey’s testing, with Anthropic’s Claude 3.7 Sonnet following close behind.
In its blog post, Harvey says that it will also now join the growing ranks of those sharing a public leaderboard of model benchmark performance. Its board will rank how major reasoning models are doing on legal tasks. And the company won’t just boil rankings down to a single number, but will also publish research where “top lawyers provide nuanced insights into model performance not captured by single-score benchmarks.”
So, not only is OpenAI-backed Harvey adopting competitors’ models, it is also amping up the pressure on its backers (including Google) to keep proving themselves. Not that OpenAI should worry much on that score. While AI benchmarking is growing increasingly complex and somewhat political, this is a world where OpenAI still shines.
“We are incredibly fortunate to have OpenAI as an investor in Harvey and key collaborator in our product,” Harvey CEO Winston Weinberg told Tech Zone Daily in a statement. “And, we are energized to add to our options for customers as we continue to serve the needs of our customers globally.”
