GoogleAudio matchupSpeech output
Gemini 3.8 Flash TTSvs.Gemini 3.8 Flash-Lite TTS
Which model is better?
Gemini 3.8 Flash TTS does better in independent tests.
Pick Gemini 3.8 Flash TTS for
Pick Gemini 3.8 Flash-Lite TTS for
Shared tests
Independent evaluators and leaderboards only, the newest shared snapshot per evaluator. With 95 % ranges a lead counts when they do not overlap; for hit rates, when it holds in a 95 % test. Without any uncertainty figure at most “slightly ahead”.
Hume AI
Gemini 3.8 Flash-Lite TTS slightly aheadGemini 3.8 Flash-Lite TTS is clearly ahead in 0 of 11 subtests and slightly in 2, Gemini 3.8 Flash TTS in none. 9 are level.
Hume Real World VoiceEQ: Extended long-form stability
as of October 7, 2026
Hume Real World VoiceEQ: Same-speaker similarity
as of October 7, 2026
Hume Real World VoiceEQ: Expressiveness
as of October 7, 2026
Show all 11 subtests
| Subtest | Gemini 3.8 Flash TTS | Gemini 3.8 Flash-Lite TTS | Verdict |
|---|---|---|---|
| Hume Real World VoiceEQ: Extended long-form stability | 3 | 3 | Gemini 3.8 Flash-Lite TTS slightly ahead |
| Hume Real World VoiceEQ: Same-speaker similarity | 4 | 4 | Gemini 3.8 Flash-Lite TTS slightly ahead |
| Hume Real World VoiceEQ: Expressiveness | 3 | 3 | level |
| Hume Real World VoiceEQ: Voice identity | 5 | 5 | level |
| Hume Real World VoiceEQ: Language stability | 3 | 3 | level |
| Hume Real World VoiceEQ: Reliability | 4 | 4 | level |
| Hume Real World VoiceEQ: Acting / Role-fit | 4 | 4 | level |
| Hume Real World VoiceEQ: Expressivity–Reliability Frontier Score | 92.0 | 91.4 | level |
| Hume Real World VoiceEQ: Acoustic quality | 3 | 3 | level |
| Hume Real World VoiceEQ: Long-form speaker stability | 3 | 3 | level |
| Hume Real World VoiceEQ: Multilingual long-form speaker stability | 3 | 3 | level |
Source · latest October 7, 2026
VoiceArena
depends on the subtestGemini 3.8 Flash TTS is clearly ahead in 1 of 8 subtests, Gemini 3.8 Flash-Lite TTS in 1. 6 are level.
VoiceArena TTS Elo: Japanese
95 % ranges 1,170 to 1,206 and 1,118 to 1,154, as of October 28, 2026
VoiceArena TTS Elo: Brazilian Portuguese
95 % ranges 1,084 to 1,110 and 1,112 to 1,142, as of October 28, 2026
VoiceArena TTS Elo: Hindi
95 % ranges 1,079 to 1,109 and 1,050 to 1,080, as of October 28, 2026
Show all 8 subtests
| Subtest | Gemini 3.8 Flash TTS | Gemini 3.8 Flash-Lite TTS | Verdict |
|---|---|---|---|
| VoiceArena TTS Elo: Japanese | 1,188 | 1,136 | Gemini 3.8 Flash TTS ahead |
| VoiceArena TTS Elo: Brazilian Portuguese | 1,097 | 1,127 | Gemini 3.8 Flash-Lite TTS ahead |
| VoiceArena TTS Elo: Hindi | 1,094 | 1,065 | level |
| VoiceArena TTS v1, US English, All, Elo | 1,059 | 1,083 | level |
| VoiceArena TTS Elo: US English | 1,059 | 1,082 | level |
| VoiceArena TTS Elo: Arabic MSA | 1,183 | 1,163 | level |
| VoiceArena TTS Elo: Vietnamese | 1,122 | 1,141 | level |
| VoiceArena TTS Elo: Mexican Spanish | 1,138 | 1,135 | level |
Source · latest October 28, 2026
Speko
depends on the subtestGemini 3.8 Flash TTS is clearly ahead in 0 of 3 subtests and slightly in 1, Gemini 3.8 Flash-Lite TTS in 0 clearly and slightly in 1. 1 is level.
Speko TTS number-reading score
as of October 17, 2026
Speko TTS robustness, weighted number/date/currency/operator correctness
as of October 17, 2026
Speko multilingual TTS blind preference Elo (Lithuanian / lt)
95 % ranges 1,669 to 1,812 and 1,710 to 1,858, as of October 17, 2026
Source · latest October 17, 2026
Artificial Analysis
Gemini 3.8 Flash TTS aheadGemini 3.8 Flash TTS is clearly ahead in 1 of 2 subtests, Gemini 3.8 Flash-Lite TTS in none. 1 is level.
Artificial Analysis Provider Voice Arena preference
95 % ranges 1,269 to 1,295 and 1,241 to 1,267, as of November 3, 2026
Artificial Analysis Controlled Voice Arena Elo
95 % ranges 1,039 to 1,063 and 1,040 to 1,064, as of November 3, 2026
Source · latest November 3, 2026
Vendor figures
Figures from vendor reports that cover both models. They do not count toward the verdict.
Show 7 vendor figures
| Figure | Source | Gemini 3.8 Flash TTS | Gemini 3.8 Flash-Lite TTS |
|---|---|---|---|
| Voice Arena TTS Elo (Arabic MSA) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,204 | 1,181 |
| Voice Arena TTS Elo (Brazilian Portuguese) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,104 | 1,134 |
| Voice Arena TTS Elo (English) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,061 | 1,087 |
| Voice Arena TTS Elo (Hindi) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,106 | 1,076 |
| Voice Arena TTS Elo (Japanese) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,232 | 1,152 |
| Voice Arena TTS Elo (Mexican Spanish) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,152 | 1,146 |
| Voice Arena TTS Elo (Vietnamese) | Google DeepMind Gemini 3.8 TTS evaluation report (Voice Arena Evals) | 1,135 | 1,156 |
All data side by side
Prices are verbatim from the provider, because audio models bill per minute, character or token.
| Gemini 3.8 Flash TTS | Gemini 3.8 Flash-Lite TTS | |
|---|---|---|
| Task | Speech output | Speech output |
| Developer | ||
| Status | Active | Active |
| Released | September 22, 2026 | September 22, 2026 |
| Runs locally | no | no |
| Price | $0.50 per 1M text tokens through December 31, 2026, then $1 per 1M text tokens from January 1, 2027 | $0.50 per 1M text tokens through December 31, 2026, then $1 per 1M text tokens from January 1, 2027 |
| Popularity (out of 100) | 0 | 0 |