TTS API comparison 2026
Cartesia Sonic 3.5 vs Deepgram Flux TTS Cartesia vs Deepgram · text-to-speech for voice agents
Cartesia Sonic 3.5 is in the Visible delay speed tier (P95 of 320 ms or more); Deepgram Flux TTS is in the Fast enough speed tier (P95 from 138 ms to 263 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.
Data read October 22, 2026 · Coval, Speko, Artificial Analysis
Side by side
| Measure | Cartesia Sonic 3.5 | Deepgram Flux TTS |
|---|---|---|
| Speed tierCoval P95, 30 days | Visible delay | Fast enough |
| Median time to first audioCoval, 30-day window | 281 ms | 200 ms |
| P95 time to first audio (slow turns)Coval, 30-day window | 381 ms | 263 ms |
| Time to first audio, board chartSpeko, US East | 132 ms | 118 ms |
| Naturalness EloSpeko | 1,574 | 1,550 |
| Listener Elo (95% interval)Artificial Analysis | 1,191 ±11 | 1,052 ±13 |
| Numbers and dates (0 to 1)Speko | 0.18 | 0.2 |
| List price per 1M charactersCoval, Speko, Artificial Analysis | $49 to $50 | $45 |
In bold: the model ahead on that measure. “tie”: latencies within 10 ms, or Elo scores whose 95% intervals overlap.
Where Cartesia Sonic 3.5 is ahead
- →Naturalness Elo: 1,574 vs 1,550 (Speko)
- →Listener Elo (95% interval): 1,191 ±11 vs 1,052 ±13 (Artificial Analysis)
Where Deepgram Flux TTS is ahead
- →Median time to first audio: 200 ms vs 281 ms (Coval, 30-day window)
- →P95 time to first audio (slow turns): 263 ms vs 381 ms (Coval, 30-day window)
- →Time to first audio, board chart: 118 ms vs 132 ms (Speko, US East)
- →Numbers and dates (0 to 1): 0.2 vs 0.18 (Speko)
- →List price per 1M characters: $45 vs $49 to $50 (Coval, Speko, Artificial Analysis)
Frequently asked questions
- What is the difference between Cartesia Sonic 3.5 and Deepgram Flux TTS?
- Cartesia Sonic 3.5 (Cartesia) is in the Visible delay speed tier (P95 of 320 ms or more); Deepgram Flux TTS (Deepgram) is in the Fast enough speed tier (P95 from 138 ms to 263 ms). Cartesia Sonic 3.5 leads on naturalness Elo and listener Elo (95% interval); Deepgram Flux TTS leads on median time to first audio, P95 time to first audio (slow turns), time to first audio, board chart, numbers and dates (0 to 1) and list price per 1M characters. No combined score is computed.
- Which is faster for real-time voice agents, Cartesia Sonic 3.5 or Deepgram Flux TTS?
- On Coval's 30-day window read on October 22, 2026, the median time to first audio is 281 ms for Cartesia Sonic 3.5 and 200 ms for Deepgram Flux TTS (Deepgram Flux TTS ahead by 81 ms); the P95, the slow turn, is 381 ms and 263 ms (Deepgram Flux TTS ahead by 118 ms). Gaps under 10 ms are ties.
- Which sounds more natural, Cartesia Sonic 3.5 or Deepgram Flux TTS?
- Speko naturalness: 1,574 for Cartesia Sonic 3.5, 1,550 for Deepgram Flux TTS. Artificial Analysis listener Elo: 1,191 ±11 and 1,052 ±13. 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.