Gradium TTS Beta
Leader Proprietary · Gradium
76.8
/100
Gradium TTS Beta is a proprietary text-to-speech (TTS) API from Gradium, measured by two independent benchmarks: Coval and Speko.
Voice agents Coval Speko
Top rated for
Best TTS for Voice agents
6 text-to-speech (TTS) models, ranked from the highest benchmark score to the lowest, out of 100.
Data checked on September 30, 2026 · updated after each benchmark capture
6 text-to-speech (TTS) models start speaking within 200 ms on 9 turns out of 10: their P90 time-to-first-audio on Coval's continuous 30-day measurement is 200 ms or less. That is the typical gap between turns in human conversation (Levinson & Torreira, 2015).
Why 200 ms at the P90
A voice agent is judged on every turn, not the average one. The P90 is the time-to-first-audio that 90% of turns stay under; a model that meets 200 ms at the P90 leaves the rest of the agent pipeline (speech recognition, language model) room to answer like a person would. Models Coval does not measure are not listed.
Alibaba Qwen3 TTS Fast
Leader Open weights · Alibaba (Qwen)
73.7
/100
Alibaba Qwen3 TTS Fast is an open-weights text-to-speech (TTS) API from Alibaba (Qwen), served by Nari Labs, measured by two independent benchmarks: Coval and Speko.
Voice agents Coval Speko
Inworld Realtime TTS-2 Flash
Strong Proprietary · Inworld
68.4
/100
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.
Spanish German French Portuguese
+13
Palabra TTS v1
Strong Proprietary · Palabra
58.6
/100
Palabra TTS v1 is a proprietary text-to-speech (TTS) API from Palabra, measured by two independent benchmarks: Coval and Speko.
Voice agents Coval Speko
Fluxions vui
Contender Proprietary · Fluxions
47
/100
Fluxions vui is a proprietary text-to-speech (TTS) API from Fluxions, measured by one independent benchmark, Coval.
Voice agents Coval
Alibaba Qwen3 TTS 1.7b
Contender Open weights · Alibaba (Qwen)
45.3
/100
Alibaba Qwen3 TTS 1.7b is an open-weights text-to-speech (TTS) API from Alibaba (Qwen), served by Baseten, measured by one independent benchmark, Coval.
Voice agents Coval
Frequently asked questions
- What is the best TTS for voice agents in 2026?
- Among the 6 models that meet the 200 ms P90 threshold, Gradium TTS Beta has the highest overall benchmark score, 76.8 out of 100, with a P50 time-to-first-audio of 47.2 ms on Coval.
- Why is a fast TTS model missing from this page?
- Either Coval does not measure it, or its P90 time-to-first-audio over 30 days is above 200 ms. Its own page shows which, with the measured values.