GPT-4o Transcribe
OpenAI's most accurate speech-to-text model, built for noisy real-world audio
About
GPT-4o Transcribe arrived March 20, 2025 as OpenAI's flagship transcription model, alongside a cheaper gpt-4o-mini-transcribe sibling and a diarization variant. OpenAI claims lower word error rates than whisper-1, with particular gains on accents, background noise, and varying speech speeds — the failure modes that matter in call centers and meeting rooms. It takes audio plus optional text (language hints, proper-noun prompts) and returns plain text. Trade-offs worth knowing: it outputs no segment-level timestamps, so subtitle/SRT workflows still need whisper-1; uploads cap at 25 MB per request; and like all API models it is a black box you cannot self-host or fine-tune. Community tooling quickly made it the default STT in bots and agents that previously used Whisper.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: lower WER than whisper-1
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: OpenAI launch March 20, 2025 corroborated by multiple outlets; Lower WER than whisper-1 is team-reported; community tooling adopted it as default STT; No segment timestamps; 25MB upload limit captured.