Copenhagen's Corti Beats OpenAI and Anthropic by 25% in Medical Coding Benchmarks

Copenhagen's Corti has released Symphony for Medical Coding, an agentic AI system that outperforms models from OpenAI, Anthropic, Amazon, Oracle, and Google by more than 25% in clinical accuracy benchmarks. The product, announced April 1 and now available through Corti's API, is purpose-built for assigning medical billing and diagnostic codes to clinical documentation.

Medical coding is the critical step between a doctor's notes and a hospital's revenue. Every diagnosis, procedure, and complication must be translated into standardized codes, primarily ICD-10 and CPT, that determine how much a provider gets paid. Human coders do this work under immense time pressure, processing dozens of records per hour. Mistakes mean lost revenue, compliance risk, or in the worst cases, missed clinical patterns. It is exactly the kind of structured, high-stakes reasoning task where specialized AI can outperform general-purpose models.

How Symphony Works

Symphony operates through a four-agent pipeline that mirrors what trained human coders do. An evidence extractor isolates conditions from clinical notes. An index navigator searches the ICD alphabetical index for candidate codes. A tabular validator checks those candidates against official coding guidelines. And a code reconciler sequences and validates the final output. Each step produces auditable reasoning, not just a predicted code.

That architecture matters. General-purpose large language models from OpenAI or Anthropic can generate plausible-looking codes, but they lack the structured reasoning chain that coding auditors require. A code needs to be defensible, not just statistically likely. Corti's approach, detailed in the research paper "Code Like Humans" accepted at EMNLP 2025, was developed using data from 5.8 million patient encounters.

Finding What Human Coders Miss

The real-world impact is already visible. In a study of Danish patient data, Symphony identified three times as many suicide attempts as human coders had documented. The cases existed in clinical notes and medication records but were missed by coders working under time pressure. That is not an edge case. It illustrates a systemic problem: manual coding consistently underreports conditions that don't directly affect billing but matter enormously for public health surveillance and patient safety.

Reach and Coverage

Corti now serves over 100 million patients annually across health systems including the UK's National Health Service. The company is headquartered in Copenhagen with offices in New York and London. Symphony supports US diagnosis coding via ICD-10-CM and procedure coding via ICD-10-PCS and CPT. European ICD-10 coding is in beta, with full support planned for the UK, Germany, France, and Denmark.

Why This Matters for European AI

The competitive positioning is deliberate. Corti is not building a general-purpose AI model and hoping healthcare is one of many applications. It is building vertical AI infrastructure specifically for clinical workflows, then benchmarking that infrastructure against the best horizontal models available. The 25% accuracy advantage over OpenAI and Anthropic suggests that domain-specific architecture still beats scale in tasks requiring structured expert reasoning.

For European healthtech, this is a significant proof point. A Copenhagen-based company with fewer than 200 employees is producing clinical AI that measurably outperforms Silicon Valley's largest labs on a task that directly affects healthcare quality and revenue. The technology runs on Corti's own infrastructure, with enterprise and sovereign cloud deployment options available, meaning European health systems can adopt it without routing patient data through US cloud providers. In a sector where data residency and regulatory compliance are non-negotiable, that matters as much as the benchmark scores.

Sources

• Danish AI lab Corti beats OpenAI and Anthropic in medical coding (Sifted, April 1, 2026)
• Corti Ships Symphony for Medical Coding (Corti, April 1, 2026)
• Corti's new Symphony AI beats OpenAI and Anthropic on medical coding (The Next Web, April 2026)