Sauti / openai
OpenAI gpt-4o-transcribe
Superseded by gpt-transcribe
WER
14.7%
95% CI 11.9%–18.3%
Rank
#12
of 12 measured
Latency
2.2s
mean, per clip
Cost
$0.36
per audio hour
10 other models are a statistical tie with this one — their confidence intervals overlap, so the ordering between them is not something this data can settle: nova-3, gpt-transcribe, Whisper large-v3 (self-hosted), universal-2, universal-3-5-pro, Whisper small (self-hosted), gpt-4o-mini-transcribe, whisper-1, scribe_v2, scribe_v1.
What it is
The flagship of the previous OpenAI generation. We benchmarked it as their best model until August 2026, which was already out of date.
Pick it when
Hard to justify today. It finished last on our board, with the widest confidence interval of any model measured, at a price two other OpenAI models beat.
Think twice when
Almost always: gpt-transcribe is better placed and cheaper, and gpt-4o-mini-transcribe is better placed and half the price. The wide interval means its exact position is uncertain, but not in a direction that rescues the value.
Assessment written against the board of 2026-08-08. The figures above are read live from the current published snapshot (2026-08-08), so if those dates differ, trust the figures.
What the vendor publishes
openai publishes no stated word error rate figure for this model on FLEURS. That absence is itself the finding. Source (read 2026-08-15).The launch post names FLEURS and states the models "consistently outperform Whisper v2 and Whisper v3 across all language evaluations", but every number lives in a chart image — no English WER is written as text anywhere on the page.
How this was measured
250 clips (2.9 hours) of unscripted conversational English from People’s Speech (MLCommons, CC-BY), human-transcribed. Scored on a held-out split (N=77) never used to tune anything, through one fixed text normalizer, with bootstrap 95% confidence intervals. One corpus, one language, one modality: batch pre-recorded. It does not tell you how this model handles your accents, your domain vocabulary, or streaming.
Cost is openai’s own published pre-recorded pay-as-you-go list price, read from their pricing page on 2026-08-08. Streaming and committed-volume rates differ.