Aller au contenu

Gradium TTS Beta Gradium · text-to-speech (TTS) API

Leader

47.2 ms P50 time-to-first-audio on Coval; ranked 1st of 112 in the TTS for Voice Agents 2026 index.

Benchmark summary: 76.8/100, ranked 1st of 112 in the TTS for Voice Agents 2026 index (Leader). Strongest criterion: real-time speed, 58.7 of 60 points.

Gradium TTS Beta is a proprietary text-to-speech (TTS) API from Gradium, measured by two independent benchmarks: Coval and Speko.

Overall score

76.8 /100

1 st of 112

Score breakdown

  • Real-time speed 58.7 /60
  • Voice quality 7.8 /20
  • Accuracy 9.8 /15
  • Price 0.5 /5

Data checked on September 30, 2026

Updated after each benchmark capture

Every point comes from a public benchmark: sources · methodology

Compare with Alibaba Qwen3 TTS Fast

Gradium TTS Beta ranks 1st of 112 in the 2026 text-to-speech (TTS) API benchmark, with a score of 76.8/100.

Real-time latency

On Coval's 30-day window ending October 1, 2026 (4,696 runs), Gradium TTS Beta has a P50 (median) time-to-first-audio (TTFA) of 47.2 ms, the lowest of the 31 models Coval measures. Its P99, the slowest 1% of turns, is 108.8 ms (the lowest of the 31 models Coval measures). Latency consistency: a standard deviation of 16.6 ms (2nd lowest of the 31 models Coval measures). Mean time to first byte (TTFB) is 50.2 ms. Speko measures a P50 TTFA of 100 ms from US East (n=30), dated September 16, 2026: 6th lowest of the 29 models Speko measures.

Accuracy (word error rate)

Coval measures an accuracy (word error rate) of 4.5% over 4,696 samples on the same window, 8th lowest of the 31 models Coval measures. Speko rates its pronunciation robustness on numbers, dates and currency at 0.28 on a 0-to-1 scale (16th highest of the 28 models Speko rates).

Voice quality

Speko rates its naturalness at 1,579 Elo from blind A/B votes (10th highest of the 41 models Speko rates).

Price

Speko lists a cost of about $57.8 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 (0.58, then 0.28): the score uses the average, 0.43. The value shown is the latest.

Benchmark measurements

Latency and accuracy — Coval (30-day window)

Mean TTFA
53.5 ms
P50 / median time-to-first-audio (TTFA)
47.2 ms
P90 TTFA
72.2 ms
P95 TTFA
86.4 ms
P99 TTFA (slowest turns)
108.8 ms
Latency consistency (standard deviation)
16.6 ms
Leading silence
3.3 ms
Time to first byte (network arrival)
50.2 ms
Accuracy: word error rate (WER)
4.5%
WER basis
pooled
Latency runs
4,696 runs
Accuracy runs
4,696 runs
Window
30d
Coval snapshot
2026-10-01T10:11:47.318752Z
Coval model ID
gradium-tts-beta-202609
Coval model page
benchmarks.coval.ai

Speko

P50 TTFA, US East
100 ms
Naturalness (Elo)
1,579
Pronunciation robustness (0 to 1)
0.28
Voice drift (lower is steadier)
11
Price per 1M characters
57.8 USD
Measured on
2026-09-16
Speko model
Gradium-TTS-Beta
Speko model page
benchmarks.speko.ai

Not collected

Similar TTS models