In the rapidly evolving world of artificial intelligence, where foundation models are the bedrock for innovation, a significant development has caught the attention of the tech community, including those interested in cryptocurrency applications leveraging AI. Harvey, a prominent player in the Legal AI space and notably backed by the OpenAI Startup Fund, has announced it will begin incorporating leading AI models from both Anthropic and Google. Why This AI Models Shift Matters This move, detailed in a recent Harvey blog post, goes beyond Harvey’s previous reliance solely on OpenAI models. While Harvey states it is not abandoning OpenAI’s technology but rather adding diversity, the integration of models from key competitors like Anthropic and Google represents a notable win for these companies in the competitive AI landscape. Harvey is a high-profile success story from the OpenAI Startup Fund’s early portfolio, having grown to a reported $3 billion valuation. The significance is amplified by Harvey’s history: Harvey was among the first four startups the OpenAI Startup Fund backed in late 2022. The company has seen rapid growth and secured substantial funding rounds. Interestingly, Google’s venture arm, GV, led Harvey’s Series C round, yet Harvey didn’t immediately adopt Google’s Google AI models until now. What Drove Harvey to Adopt New AI Models? The decision wasn’t arbitrary. Harvey developed an internal benchmark called ‘BigLaw’ to evaluate how different foundation models perform on various legal tasks. The results were compelling: A variety of models are becoming increasingly capable in legal applications. Specific models demonstrate particular strengths in different legal tasks. Within a year, multiple non-OpenAI models outperformed Harvey’s original benchmarked system. Instead of focusing resources on training its own models from scratch, Harvey concluded it was more efficient to leverage high-performing, general reasoning models from other vendors and fine-tune them specifically for the legal market. This includes accessing models from Anthropic and Google, sometimes via cloud providers like Amazon. Benchmarking Reveals Task-Specific Strengths Harvey’s BigLaw benchmark provided crucial insights into which AI models excel at what. For example, their testing indicated: Google AI ‘s Gemini 2.5 Pro shows strength in legal drafting. However, Gemini 2.5 Pro reportedly struggles with certain pre-trial tasks, like writing oral arguments, due to challenges understanding complex evidentiary rules such as hearsay. OpenAI’s models perform well on such pre-trial tasks. Anthropic ‘s Claude 3.7 Sonnet also performs strongly in these areas, following closely behind OpenAI’s performance in some pre-trial tasks. These findings underscore the reality that no single model is universally superior for all tasks, prompting the need for a multi-model strategy in specialized fields like Legal AI . Future Implications: AI Agents and Transparency Using a diverse set of AI models will also be beneficial as Harvey develops AI agents for legal workflows. Furthermore, Harvey plans to contribute to the broader AI community by sharing its BigLaw benchmark findings publicly. They intend to launch a public leaderboard ranking major reasoning models on legal tasks. This leaderboard won’t just rely on single scores but will also incorporate qualitative insights from experienced lawyers, offering a more nuanced view of model performance. This move not only diversifies Harvey’s technological foundation but also puts pressure on its model providers, including its investor OpenAI and investor Google (via GV), to continue innovating and proving their models’ capabilities in real-world, high-stakes applications like law. Harvey’s Perspective and the Competitive Landscape Despite integrating competitor models, Harvey maintains a positive relationship with its initial backer. Harvey CEO Winston Weinberg stated, “We are incredibly fortunate to have OpenAI as an investor in Harvey and key collaborator in our product. And, we are energized to add to our options for customers as we continue to serve the needs of our customers globally.” This strategic shift by a leading Legal AI company like Harvey highlights the intense competition among foundation model providers. While OpenAI remains a dominant force, companies like Anthropic and Google are rapidly advancing their capabilities, making multi-model strategies increasingly attractive for developers building specialized AI applications. The focus is shifting towards leveraging the best available tool for each specific job, rather than being locked into a single provider, even when that provider is also an investor. To learn more about the latest AI models trends, explore our article on key developments shaping AI features.
