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 was initially built exclusively on OpenAI's models and was backed by the OpenAI Startup Fund 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 runs a multi-model stack that includes Claude alongside models from OpenAI, Google, and Mistral (launched in May 2026 as an opt-in early access option for eligible EU customers only; Harvey said US and Australian availability would be announced later), choosing among them based on the legal task at hand. Per Harvey's own product updates, Claude Fable 5 was added on an opt-in basis in June 2026, and Claude Opus 5 began rolling out to US customers in July 2026, with Harvey saying EU and AU availability would follow. Claude Fable 5.1 followed on September 1, 2026, also as an opt-in model. At the May 2025 launch of its multi-model architecture, according to Harvey and TechCrunch's reporting, the initial lineup included:

• OpenAI: including GPT-4.1 and the o3 reasoning model. Harvey found OpenAI's o3 particularly strong on pre-trial tasks.
• Anthropic Claude: Claude 3.7 Sonnet at the time, which TechCrunch reported followed closely behind o3 on pre-trial tasks.
• Google Gemini: Gemini 2.5 Pro at the time, which Harvey found especially strong at drafting.

Those specific model versions reflect the 2025 launch; the providers have since shipped newer generations, and Harvey has kept adding them (for example, OpenAI's GPT-5.6 Sol went live for US customers in July 2026).

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 aims to route work to the model it judges 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, Harvey said it would integrate the new models through their cloud providers (Anthropic's Claude via AWS Bedrock and Google's Gemini via Vertex AI) with the same security and privacy guarantees as the rest of its platform. Harvey also states that it does not use inputs, outputs, or uploaded documents to train underlying models. The opt-in Fable models come with a caveat: Harvey says using them may involve data processing practices that differ from the data commitments in its existing customer agreements, and because Anthropic offers no regional processing for Fable, all Fable data is processed in the United States.

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 was part of Anthropic's Model Context Protocol (MCP) program announcement, which its CEO said "means easier access to Harvey," 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. Legal IT Insider noted that Harvey, though funded by OpenAI, has long been a partner of Anthropic, and Harvey CEO Winston Weinberg told the publication, "We integrate with everything they release." Claude is now both one of the models inside Harvey and a way to reach Harvey's workflows.

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. Depending on the task and Harvey's routing, the underlying reasoning may come from Claude or from an OpenAI or Google model; Harvey has not disclosed how often each is used. 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 Claude, one of the models Harvey draws on, is available to you directly, a solo or small firm may be able to get much of the same underlying capability by using Claude itself. Claude Team standard seats run $20 per user per month billed annually, or $25 billed monthly (see our Claude pricing guide for lawyers). Harvey does not publish pricing, but one buyer at a large firm reported on Reddit that Harvey initially quoted $1,200 per user per month ($2,400 with Lexis integration), then later offered a lower tier around $399. Against that one reported $1,200 quote, the per-seat difference is roughly 50 to 60 times, 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.