Does Harvey AI Use Claude? The Models Behind Harvey (2026)

Yes. As of 2026, Harvey AI uses Anthropic's Claude models as part of a multi-model architecture. Harvey did not start there — it launched as an OpenAI-exclusive product, backed by the OpenAI Startup Fund after being founded in 2022. But on May 13, 2025, Harvey announced it was adding Anthropic (Claude) and Google (Gemini) models to its platform. In Harvey's own words: "The first change we are making is to incorporate and optimize leading foundation models from Anthropic and Google for use across the Harvey platform." So if you are asking which AI Harvey uses, the accurate answer is: several, and Claude is one of them.

What Models Does Harvey AI Use?

Harvey is not built on a single large language model. It draws on a mix of frontier models from three providers, choosing among them based on the legal task at hand. According to Harvey and TechCrunch's reporting, the lineup includes:

• OpenAI — including GPT-4.1 and the o3 reasoning model. Harvey found OpenAI's o3 particularly strong on pre-trial tasks.
• Anthropic Claude — Claude Sonnet 3.7, which Harvey rated close behind the leaders on complex legal tasks.
• Google Gemini — Gemini 2.5 Pro, which Harvey found especially strong at drafting.

The point of going multi-model is that no single model wins at everything. Different models have different strengths across drafting, reasoning, and analysis — so Harvey routes work to whichever model performs best for a given job. For a deeper look at how the leading models compare for legal work, see our guide on which Claude model lawyers should use.

How Harvey Picks a Model

Harvey handles model selection in two ways. By default, it uses auto-routing: the platform automatically picks the model it judges best for each task, so users don't have to think about it. For users who want control, Harvey also offers a manual model selector to choose a specific model. Behind the scenes, the models are served through the providers' own clouds — Anthropic's Claude via AWS Bedrock and Google's Gemini via Vertex AI — all running under Harvey's existing security and privacy guarantees rather than being retrained on customer data.

The Harvey–Anthropic Relationship Is Deepening

Adding Claude in 2025 was the start, not the end, of Harvey's relationship with Anthropic. In February 2026, the two companies moved closer together in a way that matters even if you never log into Harvey. According to Legal IT Insider, Harvey became accessible through Anthropic's Model Context Protocol (MCP) program, and Harvey's legal workflows became invokable from within Claude itself through a connector announced that month.

In practical terms, that means Claude can act as the orchestrator of a multi-step legal workflow while Harvey supplies the domain-specific legal capability. Harvey CEO Winston Weinberg described Harvey as "a long partner of Anthropic's" and said the company aims to "integrate with everything they release." The direction of travel is clear: Claude is increasingly central to how Harvey works.

What This Means for Solo and Small Firms

Here is the practical takeaway, and it is an important one. Harvey is a legal-specific workflow and agent layer that sits on top of general-purpose foundation models — including Claude. The underlying reasoning that powers many of Harvey's tasks comes from models like Claude Sonnet. Harvey's value is the legal-specific scaffolding around those models: vetted workflows, integrations, and an enterprise wrapper.

That distinction has real cost implications. Because the reasoning engine for many tasks is a model you can already access directly, a solo or small firm can get much of the same underlying capability by using Claude itself. Claude Team runs roughly $30 per user per month. Harvey, by contrast, is reported to cost around $1,200 per seat per month — with a reported entry point near $288,000 per year and no free trial. That is roughly a 40x difference in per-seat cost for access to overlapping core technology.

None of this is a knock on Harvey. It is a serious product with serious adoption: it reached an $11 billion valuation in March 2026, and reports more than 100,000 lawyers and a majority of the Am Law 100 as users. For large firms that need its enterprise integrations, governance, and pre-built workflows, that price can be justified. But if you are a solo practitioner, a small firm, or in-house counsel doing contract review, drafting, and research, you may not need the full workflow layer to capture most of the benefit.

If you want to weigh the two side by side, we break it down in Claude vs. Harvey for lawyers, the details of Harvey's pricing, our full Harvey AI review, and a roundup of Harvey alternatives. To get started with the underlying model directly, see how to use Claude for legal work. In-house teams can also explore our resources for corporate counsel.