AssemblyAI's Universal-3.6 Pro speech-to-text model has launched on OpenRouter via Sync API, with 30% fewer errors on short conversational utterances and 12% fewer average errors across supported languages compared to Universal-3.5 Pro, plus 14 new languages and improved recognition of phone numbers, addresses, emails, and code.
AssemblyAI's Universal-3.5 Pro speech-to-text model has launched on OpenRouter, supporting 19 languages in a single synchronous call, ranking first in accuracy across independent benchmarks, handling audio up to two minutes with word-level timestamps and confidence scores, and offering a 50% discount through September 29.