69.4 ms P50 time-to-first-audio on Coval; ranked 5th of 112 in the TTS for Voice Agents 2026 index.
Benchmark summary: 68.4/100, ranked 5th of 112 in the TTS for Voice Agents 2026 index (Strong). Strongest criterion: price, 4 of 5 points.
Inworld Realtime TTS-2 Flash is a proprietary text-to-speech (TTS) API from Inworld, measured by three independent benchmarks: Artificial Analysis, Coval and Speko.
Overall score
68.4 /100
Data checked on September 30, 2026
Updated after each benchmark capture
Every point comes from a public benchmark: sources · methodology
Inworld Realtime TTS-2 Flash ranks 5th of 112 in the 2026 text-to-speech (TTS) API benchmark, with a score of 68.4/100.
On Coval's 30-day window ending October 1, 2026 (11,660 runs), Inworld Realtime TTS-2 Flash has a P50 (median) time-to-first-audio (TTFA) of 69.4 ms, 4th lowest of the 31 models Coval measures. Its P99, the slowest 1% of turns, is 175.2 ms (5th lowest of the 31 models Coval measures). Latency consistency: a standard deviation of 373.3 ms (25th lowest of the 31 models Coval measures). Mean time to first byte (TTFB) is 75.7 ms. Speko measures a P50 TTFA of 82 ms from US East (n=30), dated September 3, 2026: 4th lowest of the 29 models Speko measures.
Coval measures an accuracy (word error rate) of 5.4% over 11,640 samples on the same window, 25th lowest of the 31 models Coval measures. Speko rates its pronunciation robustness on numbers, dates and currency at 0.58 on a 0-to-1 scale (7th highest of the 28 models Speko rates).
In the Artificial Analysis Speech Arena (capture of October 1, 2026), Inworld Realtime TTS-2 Flash has an Elo of 1,209 (±15, 95% confidence) from 1,943 blind comparisons, 9th highest of the 92 models Artificial Analysis rates. Speko rates its naturalness at 1,501 Elo from blind A/B votes (25th highest of the 41 models Speko rates).
Artificial Analysis lists Inworld Realtime TTS-2 Flash at $10.4 per 1M characters (14th lowest of the 78 models Artificial Analysis lists a price for). Speko lists a cost of about $15 per 1M characters.
Artificial Analysis changed its Speech Arena Elo between the captures of September 30, 2026 and October 1, 2026 without a new measurement date (1,210, then 1,209): the score uses the average, 1,209.5. The value shown is the latest. Speko changed its pronunciation robustness between the captures of September 30, 2026 and October 1, 2026 without a new measurement date (0.93, then 0.58): the score uses the average, 0.755. The value shown is the latest.
Tested in Spanish, German, French, Portuguese, Japanese, Chinese, Arabic, Hindi and Vietnamese by Artificial Analysis or Speko. A language missing here was not tested, which does not mean it is unsupported.
Inworld Realtime TTS-2 is a proprietary text-to-speech (TTS) API from Inworld, measured by three independent benchmarks: Artificial Analysis, Coval and Speko.
Gradium TTS (Aug 2026) is a proprietary text-to-speech (TTS) API from Gradium, measured by three independent benchmarks: Artificial Analysis, Coval and Speko.
Palabra TTS v1 is a proprietary text-to-speech (TTS) API from Palabra, measured by two independent benchmarks: Coval and Speko.