Introducing Melia, our new multilingual speech-to-text model

Speechmatics launched Melia, a multilingual speech-to-text model supporting code-switching across 55+ languages, now available in production preview for batch transcription starting at $0.129/hour. The model outperforms competitors on most FLEURS benchmark languages, with a 91% win rate against Deepgram and Microsoft and 77% against AssemblyAI, and delivers 5% lower error rates than Standard on challenging monolingual audio. Melia runs alongside the existing Standard and Enhanced models rather than replacing them, targeting contact center analytics, broadcast captioning, and compliance monitoring.