He Tool He Popular Legal Harvey will now use the main models of the Foundation from Anthropic and Google, moving beyond using Openai strictly, Harvey announced in a blog post on Tuesday.
This is important because Harvey is one of the most successful portfolio companies supported at the beginning of OpenAi. The Openai Start Fund is a OpenAi -related fund to support companies that develop products at the top of it, mainly of themselves. While Harvey says he is not abandoning Openai, simply adding more models and clouds, this is still a big coup for the great Openai competitors.
Harvey is one of the first four beginnings supported by the Openai starting fund, said in December 2022. This returned when Openai Sam Altman’s CEO was still running the fund. (Others in that first group include description, MEM and talk.)
Harvey, who has grown as crazy since that time, is now a $ 3 billion rating start, said in February, when he announced a 300 million dollars Dollars led by Sequoia, with other big names such as Coatue, Kleiner Perkins, and Openai Poly in POLING IN.
Interesting, Google’s Entrepreneurship Wing, GV, led Harvey 100 million series in July 2024 (and Openi Fund participated in that round as well). But Harvey did not immediately approve Google’s models after setting up Google Corporate Entrepreneurship firm on its lids table. (GV also participated in D Harvey Series.)
So what persuaded Harvey to go beyond Openai models now? The developed internal starting order, called Biglaw, showed that a variety of basic models are growing more and more capable of a variety of legal tasks and some are better in specific tasks than others.
Instead of spending training models of her efforts, Harvey realized, he could simply embrace high -performance foundation models, reasoning from other retailers (eg Google and Anthropic through Amazon’s Cloud) and then fix them for the legal market.
Using a variety of models will also help as Harvey creates agents, the company says.
“In less than a year, seven models (including three non-oai models) now exceed the Harvey system originally standardized in Biglaw Bench,” Harvey wrote in the blog post.
Harvey’s standard also showed that different foundation models are better in specific legal tasks than others. For example, she says Google Gemini 2.5 pro “excels” in legal drafting, but “fights” with preliminary tasks such as writing oral arguments because the model does not fully understand the “complex rules of testimony like Hearsay”.
Openi’s O3 does such prejudicial tasks, according to Harvey’s test, with anthropic Claude 3.7 follow -up.
In her blog post, Harvey says he will also join the growing ranks of those who share a public model performance leader. His board will list how the main patterns of reasoning in legal duties are doing. And the company will not only lead to the deduction rankings to a single number, but will also publish research where “the main lawyers offer nuanced knowledge in model performance that is not captured by standards with one results.”
So not only Harvey backed by Openai is adopting competitors’ models, but it is also increasing pressure on its supporters (including Google) to continue to try themselves. Not that Openai has to worry a lot about that result. While benchmarking he is growing more and more complex and somewhat political, this is a world where Openai still shines.
“We are very lucky to have Openai as an investor in Harvey and the main associate in our product,” CEO Harvey Winston Weinberg told Techcrunch in a statement. “And we are energy to add our customers options as we continue to serve the needs of our clients globally.”
