Anthropic, Google score win by nabbing OpenAI-backed Harvey as a user

Harvey announced in a blog post on Tuesday that the famous AI’s legal tool will now use the leading basic models of anthropor and Google, and beyond precisely using Openai’s.

This is noticeable because Harvey is one of the most wallet companies supported by the Openai startup Fund early. The Openai Startup Fund is a box -related box to support companies that develop products at the top of artificial intelligence technologies, especially Openai. While Harvey says he does not abandon Openai, only adding more models and clouds, this is still a major coup for big competitors in Openai.

Harvey, one of the first four startups supported by Openai, said in December 2022. It was when Openai CEO was still running the box. (Includes others in that first regiment Decript and MEM and Speak.)

Harvey, who has grown like MAD since then, is now 3 billion dollars in February, when it announced the D 300 million dollar series led by Sequoia, with other big names such as Coature, Kleiner Perkins and Openai Fund Fund.

Interestingly, the Google Adventure arm, GV, led the Harvey series of $ 100 million in July 2024 (and the Openai Fund also participated in that round as well). But Harvey did not immediately adopt the embroidery of IQ from Google after the Google Projects Company has placed on its maximum schedule. (Gv participated in the D Harvey series as well.)

So, what convinced Harvey to overcome Openai models now? The internally advanced standard for starting start, called Biglaw, has shown that a wide range of basic models are increasingly increasing in a set of legal tasks and some are better in specific tasks than others.

Instead of eliminating her efforts in training models, Harvey determines that he can simply adopt the models of the Higher Thinking Foundation from other sellers (for example Google and Anthropor via Cloud’s Cloud) and then control them in the legal market.

The company says that using a variety of models will also help as Harvey creates artificial intelligence agents.

“In less than a year, it is now outperforming seven models (including three non -OAI models) now on the Harvey system originally based on Biglaw Bench”.

The Harvey Index also showed that different basic models are better in specific legal tasks than others. For example, “Gemini 2.5 Pro” from Google “Excels” says in legal formulation but “struggles” with pre -trial tasks such as writing oral arguments because the model does not fully understand “complex proof rules such as rumors.”

Openai’s O3 performs such previous tasks to trial well, according to Harvey’s test, with Claude 3.7 Sonnet followers of man.

In its blog post, Harvey says it will now join the growing ranks of those who participate in the performance of the standard performance. Its council will classify how the main thinking models are based on legal tasks. The company will not be summarized in the classification to one, but it will also publish research as “senior lawyers offer accurate visions of the typical performance that is not captured by unilateral standards.”

Therefore, not only Harvey supported by Openai adopts competitors’ models, but it increases the pressure on its supporters (including Google) to continue to prove themselves. Openai should not worry a lot about this result. While artificial intelligence standards grow in a fairly complex and political, this is a world where Openai is still lighting.

“We are incredibly lucky to be on Openai as an investor in Harvey and the main collaborator in our products,” Winston Winston Winberg, CEO of Harvey, told Techcrunch in a statement. “We are vibrant to add our options to customers as we continue to meet the needs of our customers worldwide.”