AssemblyAI

AssemblyAI's Sync API runs Universal-3.5 Pro on clips up to two minutes and returns the finished transcript in the same HTTP response, about 134 ms at p50 by AssemblyAI's figures. It launched in July 2026 for dictation and push-to-talk.

Universal-3.5 Pro is AssemblyAI's current flagship pre-recorded model, released July 2026, transcribing 18 languages with mid-sentence code switching. It predicts speaker labels jointly with the words and accepts a prompt describing the audio's domain.

Streaming version of Universal-3.5 Pro, released June 2026, that uses the voice agent's last question and earlier turns as context and offers min_latency, balanced and max_accuracy modes. Universal-3.6 Pro Realtime superseded it in September 2026.

Hosted by AssemblyAI, this streaming version of OpenAI's open-source Whisper (API id whisper-rt) arrived in March 2026 with 99+ languages and automatic language detection. It no longer appears in AssemblyAI's current streaming documentation.

Universal-3 Pro was AssemblyAI's first promptable batch model, launched February 2026, accepting natural-language prompts of up to 1,500 words and up to 1,000 keyterms, with code switching across six languages. Deprecated in July 2026 in favour of Universal-3.5 Pro.

Low-cost multilingual counterpart to Universal-Streaming English, released October 2025 for English, Spanish, French, German, Italian and Portuguese. Language is detected per turn rather than mid-sentence, with a language code returned for each utterance.

English-only streaming model from June 2025, built for voice agents with about 300 ms word emission, immutable transcripts and end-of-turn detection. AssemblyAI now positions it as a low-cost tier and recommends Universal-3.6 Pro Realtime for new agents.

Universal-2 is AssemblyAI's October 2024 batch speech-to-text model covering 99 languages. Superseded for top accuracy by Universal-3.5 Pro, it remains the lower-cost tier and the automatic fallback for languages Universal-3.5 Pro does not support.