# Harvey embraces open AI models to reduce reliance on Anthropic and OpenAI > Source: > Published: 2026-09-21 16:34:31+00:00 Photo: Tima Miroshnichenko / Pexels # Harvey embraces open AI models to reduce reliance on Anthropic and OpenAI The legal AI startup's gross margins cratered from 50% to -50% in six months, forcing a hard rethink of its model strategy Harvey built its legal AI platform on the backs of OpenAI and Anthropic. Then the bill arrived. The company, which provides AI tools to law firms and corporate legal departments, watched its gross margins collapse from roughly 50% to negative 50% within six months of a March 2026 product update that triggered a sharp surge in customer usage. More customers, more queries, more inference costs from proprietary providers. The math stopped working fast. Harvey’s response: build its own model and reduce the power that external AI providers hold over its cost structure. The company launched **Tenet**, its first internally post-trained open-weight model, around August 20, 2026. It is built on Kimi K3, an open-weight base model with approximately 2.8 trillion parameters that was itself open-sourced around July 2026. ## From cost crisis to custom model Harvey apparently moved quickly. Going multi-model is not new for the company: it integrated Anthropic and [Google](https://cryptobriefing.com/markets/alphabet/) models alongside OpenAI back in May 2025, creating a layered architecture that could route tasks to different providers. But routing among expensive proprietary models only helps so much when all of them charge meaningfully for inference at scale. Kimi K3 changes the calculation. Open-weight models can be deployed on infrastructure that Harvey controls, which means inference costs become a function of compute spending rather than per-token API fees. The company’s platform now evaluates models using benchmarks tailored to legal work, including BigLaw Bench, optimizing which model handles which task based on performance and cost. Tenet is Harvey’s bet that fine-tuning an open-weight foundation on legal data produces results competitive with larger proprietary models for domain-specific work. The company describes it as a state-of-the-art solution for legal tasks. ### AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. ## Valuation keeps climbing despite the turbulence What makes Harvey’s margin crisis notable is that it happened during a period of explosive growth, not decline. The company’s valuation jumped from $11 billion in March 2026 to $15.5 billion by September 2026, when it closed a $550 million funding round. Annual recurring revenue surpassed $400 million, and the platform now serves more than 3,000 customers. ## A template for the broader legal tech industry Legal work is a useful testing ground for this kind of model strategy. It is highly document-intensive, which means inference volumes scale sharply with client activity. It also demands precision: a hallucination in a contract review or case brief is not a minor inconvenience, it is a professional liability. That combination of high volume and high accuracy requirements makes inference costs a central strategic problem much faster than in lower-stakes applications. The Kimi K3 base, with its 2.8 trillion parameters, offers Harvey a foundation large enough to be competitive while open enough to be customized. The legal-specific post-training that produced Tenet is where Harvey’s proprietary value sits. The underlying weights are not secret, but the fine-tuning, the benchmark evaluation pipeline, and the task-routing architecture are harder to replicate. The Tenet launch also signals something about where Harvey’s competitive moat is being repositioned. The company spent its early years competing on the quality of its legal-specific application layer, essentially the interface and workflow built on top of foundation models anyone could access. Tenet represents a move toward competing on the model layer itself, a significantly more capital-intensive and technically demanding position to hold, but one that reduces dependence on providers who are also, increasingly, potential competitors. With $550 million in fresh capital, Harvey has the runway to find out whether that bet pays off. **Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).