Google: Gemini 3.5 TranscribeNEW

google/gemini-3.5-transcribe

transcriptionAudio

Google's speech-to-text model: synchronous transcription with word-level timestamps and speaker diarization — up to eight speakers in one track. Audio up to one hour (30 minutes with timestamps or diarization). Served via /v1/audio/transcriptions; billed by actual input audio tokens.

from openai import OpenAI

client = OpenAI(base_url="https://api.mixen.ai/v1", api_key="mxn-...")
tr = client.audio.transcriptions.create(
    model="google/gemini-3.5-transcribe",
    file=open("audio.mp3", "rb"),
)
print(tr.text)

Related models

Whisper Large v3 Turbo

Voxtral Mini Transcribe

Nemotron 3.5 ASR Streaming 0.6B

MAI-Transcribe 2

GPT-4o Transcribe

GPT-4o Mini Transcribe