OpenAI GPT-4o mini TTS OpenAI · text-to-speech (TTS) API

Real-time latency

On Coval's 30-day window ending October 1, 2026 (11,780 runs), OpenAI GPT-4o mini TTS has a P50 (median) time-to-first-audio (TTFA) of 567.4 ms, 30th lowest of the 31 models Coval measures. Its P99, the slowest 1% of turns, is 5,896.6 ms (30th lowest of the 31 models Coval measures). Latency consistency: a standard deviation of 1,575.6 ms (30th lowest of the 31 models Coval measures). Mean time to first byte (TTFB) is 962.7 ms. Speko measures a P50 TTFA of 691 ms from US East (n=30), dated July 3, 2026: 26th lowest of the 29 models Speko measures.

Accuracy (word error rate)

Coval measures an accuracy (word error rate) of 4.7% over 11,801 samples on the same window, 12th lowest of the 31 models Coval measures. Speko rates its pronunciation robustness on numbers, dates and currency at 0.44 on a 0-to-1 scale (10th highest of the 28 models Speko rates).

Voice quality

Speko rates its naturalness at 1,424 Elo from blind A/B votes (37th highest of the 41 models Speko rates).

Price

Speko lists a cost of about $20 per 1M characters.

Scoring note

Speko changed its pronunciation robustness between the captures of September 30, 2026 and October 1, 2026 without a new measurement date (0.93, then 0.44): the score uses the average, 0.685. The value shown is the latest.

Languages tested

Tested in Spanish, German, French, Japanese, Chinese, Korean, Filipino, Norwegian and Thai by Artificial Analysis or Speko. A language missing here was not tested, which does not mean it is unsupported.