TTS API comparison 2026
Gradium TTS vs ElevenLabs Eleven v4 Turbo Gradium vs ElevenLabs · text-to-speech for voice agents
Gradium TTS is in the Headroom speed tier (P95 under 120 ms); ElevenLabs Eleven v4 Turbo is in the Fast enough speed tier (P95 from 120 ms to 293 ms). Below, each measure side by side with its source; a model leads on a measure only beyond the tie rule. No combined score, no overall winner.
Figures as of October 27, 2026 · Coval, Speko, Artificial Analysis, Cekura
Side by side
| Measure | Gradium TTS | ElevenLabs Eleven v4 Turbo |
|---|---|---|
| Speed tierCoval P95, 30 days | Headroom | Fast enough |
| Median time to first audioCoval, 30-day window | 54 ms | 189 ms |
| P95 time to first audio (slow turns)Coval, 30-day window | 104 ms | 239 ms |
| Time to first audio, board chartSpeko, US East | 55 ms | 203 ms |
| Time to first audio, full sentence (median)Cekura | 76 ms | 176 ms |
| Time to first audio, text streamed word by word (median)Cekura | -64 ms | 69 ms |
| Naturalness EloSpeko | 1,592 | 1,670 |
| Listener Elo (95% interval)Artificial Analysis | 1,174 ±13 | 1,333 ±16 |
| Numbers and dates (0 to 1)Speko | 0.27 | 0.9 |
| Hard passages pronounced correctly (95% interval)Artificial Analysis, human reviewers | not tested | 90.1% (87.8 to 92.3) |
| Reading accuracy on hard textCekura, transcriber-scored | 79.5% | 79.9% |
| List price per 1M charactersCoval, Speko, Artificial Analysis, Cekura | $36 to $57.8 | $40 |
In bold: the model ahead on that measure. “tie”: latencies within 10 ms, or Elo scores whose 95% intervals overlap.
Where Gradium TTS is ahead
- →Median time to first audio: 54 ms vs 189 ms (Coval, 30-day window)
- →P95 time to first audio (slow turns): 104 ms vs 239 ms (Coval, 30-day window)
- →Time to first audio, board chart: 55 ms vs 203 ms (Speko, US East)
- →Time to first audio, full sentence (median): 76 ms vs 176 ms (Cekura)
- →Time to first audio, text streamed word by word (median): -64 ms vs 69 ms (Cekura)
- →List price per 1M characters: $36 to $57.8 vs $40 (Coval, Speko, Artificial Analysis, Cekura)
Where ElevenLabs Eleven v4 Turbo is ahead
- →Naturalness Elo: 1,670 vs 1,592 (Speko)
- →Listener Elo (95% interval): 1,333 ±16 vs 1,174 ±13 (Artificial Analysis)
- →Numbers and dates (0 to 1): 0.9 vs 0.27 (Speko)
- →Reading accuracy on hard text: 79.9% vs 79.5% (Cekura, transcriber-scored)
Frequently asked questions
- What is the difference between Gradium TTS and ElevenLabs Eleven v4 Turbo?
- Gradium TTS (Gradium) is in the Headroom speed tier (P95 under 120 ms); ElevenLabs Eleven v4 Turbo (ElevenLabs) is in the Fast enough speed tier (P95 from 120 ms to 293 ms). Gradium TTS leads on median time to first audio, P95 time to first audio (slow turns), time to first audio, board chart, time to first audio, full sentence (median), time to first audio, text streamed word by word (median) and list price per 1M characters; ElevenLabs Eleven v4 Turbo leads on naturalness Elo, listener Elo (95% interval), numbers and dates (0 to 1) and reading accuracy on hard text. No combined score is computed.
- Which is faster for real-time voice agents, Gradium TTS or ElevenLabs Eleven v4 Turbo?
- On Coval's 30-day window as of October 27, 2026, the median time to first audio is 54 ms for Gradium TTS and 189 ms for ElevenLabs Eleven v4 Turbo (Gradium TTS ahead by 135 ms); the P95, the slow turn, is 104 ms and 239 ms (Gradium TTS ahead by 134 ms). Gaps under 10 ms are ties.
- Which sounds more natural, Gradium TTS or ElevenLabs Eleven v4 Turbo?
- Speko naturalness: 1,592 for Gradium TTS, 1,670 for ElevenLabs Eleven v4 Turbo. Artificial Analysis listener Elo: 1,174 ±13 and 1,333 ±16. The two sources use different listeners and can disagree.
- How are these two text-to-speech APIs compared?
- Speed first, on Coval's median and P95 time to first audio; then voice quality (Speko, Artificial Analysis), accuracy on numbers and dates (Speko) and list price. Latencies within 10 ms and Elo scores whose intervals overlap are ties. Scores from different benchmarks are never added into one number.
- Can these numbers be checked?
- Yes. Each model page lists every measurement with its source and the date of the reading, and links to the benchmark page it comes from. The rules are on the methodology page.