Voxtral Transcribe 2 consists of two speech-to-text models with transcription quality, diarization, and ultra-low latency.
Traces of two new AI models, codenamed Avocado, were reportedly spotted in the codebase of the Meta AI platform.