Solaria-3

Gladia's most accurate speech-to-text model, best-in-class for business audio in core European languages (EN/FR/DE/ES/IT). #1 on Earnings22 (6.4% WER) and Switchboard; tuned for noisy, fast-paced conversational calls with accents, company names, and jargon. Async batch transcription with diarization, translation, summarization, sentiment, NER, PII redaction, chapterization, and audio-to-LLM. Complements Solaria-1 (which keeps the edge on language breadth and code-switching).