product · May 17, 2026

Together AI Highlights NVIDIA Parakeet TDT 0.6B V3 Speech-to-Text Performance

Together AI highlighted its speech-to-text models holding the top two positions for transcription speed on the Artificial Analysis Speech to Text leaderboard in a recent LinkedIn post. The NVIDIA Parakeet TDT 0.6B V3 model hosted on Together AI transcribes 303 seconds of audio per second of processing time with $1.50 per 1,000 minutes of audio pricing and 4.6% AA-WER across three real-world datasets. Together AI positions fast low-cost STT as core infrastructure for AI natives building real-time voice agents on its AI Native Cloud. The performance and cost profile strengthens Together AI's stance in latency-sensitive AI applications.