Speechmatics Achieves 93% Accuracy in Medical Speech-to-Text, Setting New Industry Standard with 50% Fewer Clinical Errors and Advanced Speaker Diarization

Speechmatics has set a new benchmark in medical speech-to-text technology with its latest model achieving 93% real-world accuracy, marking a significant leap in clinical transcription. The enhanced system delivers 50% fewer keyword errors on medical terms and 17% lower overall word errors compared to the next best competitor, demonstrating superior performance in high-stakes healthcare environments. Designed for the fast-paced, complex nature of clinical settings, the new model features expanded medical vocabulary, including drug names, dosages, procedures, and precise numerical and temporal formatting. It also supports accent-independent recognition and real-time speaker diarization, accurately distinguishing between clinicians, patients, and family members—even in noisy or interrupted conversations. The result is cleaner, more accurate clinical notes with fewer manual corrections, leading to better documentation, improved patient handovers, and reduced administrative burden on healthcare teams. By minimizing errors where they matter most—such as diagnoses, medications, and treatment timelines—this technology helps ensure patient safety and enhances clinician efficiency. Katy Wigdahl, CEO of Speechmatics, said: “Our goal is simple: build speech technology clinicians can trust in the messiness of real-world practice. When every voice is understood, the experience feels human again. Every accurate word returns time to patients and eases clinician burnout. This is why we build—so teams can focus on what matters.” The model is available in both batch and real-time modes, making it ideal for AI-Scribe and dictation-driven workflows. Its real-time-first architecture ensures that performance doesn’t degrade when switching from file-based to live transcription—a key advantage over many competitors. In independent benchmarking, Speechmatics achieved a Keyword Error Rate (KER) of just 4%, outperforming all evaluated systems. The next best model scored 91% accuracy, while the broader peer range spanned 74% to 91%. These results confirm Speechmatics’ leadership in clinical-grade speech recognition. The new model is particularly well-suited for the growing trend toward ambient clinical documentation, where AI captures and transcribes conversations automatically during patient visits. With robust performance across diverse accents, speaking styles, and complex multi-speaker interactions, it supports scalable, accurate documentation across hospitals, clinics, and telehealth platforms. Speechmatics’ technology is built to work in the real world—handling overlapping speech, background noise, and rapid turn-taking with precision. The company’s Speech-to-Text API supports over 55 languages, adapts to regional accents, and offers flexible deployment across cloud, on-premise, and edge environments. Trusted by global enterprises across healthcare, media, contact centers, and AI-driven workflows, Speechmatics powers voice experiences that are accurate, reliable, and scalable. Headquartered in Cambridge and London, the company continues to advance the state of speech recognition for mission-critical applications.