DeepgramCartesiaAudio matchupSpeech output
Aura-2vs.Sonic 3.6
Which model is better?
Aura-2 is slightly ahead in independent tests, but the lead is not certain.
Pick Aura-2 for
Pick Sonic 3.6 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”.
Coval
depends on the subtestSonic 3.6 is clearly ahead in 0 of 12 subtests and slightly in 5, Aura-2 in 0 clearly and slightly in 4. 3 are level.
Coval TTFA roundtrip component maximum
lower is better, as of November 5, 2026
Coval Time to First Audio (TTFA) maximum
lower is better, as of November 5, 2026
Coval TTFA leading silence maximum
lower is better, as of November 5, 2026
Show all 12 subtests
| Subtest | Aura-2 | Sonic 3.6 | Verdict |
|---|---|---|---|
| Coval TTFA roundtrip component maximum (lower is better) | 464 ms | 2,932 ms | Aura-2 slightly ahead |
| Coval Time to First Audio (TTFA) maximum (lower is better) | 808 ms | 3,066 ms | Aura-2 slightly ahead |
| Coval TTFA leading silence maximum (lower is better) | 534 ms | 245 ms | Sonic 3.6 slightly ahead |
| Coval Word Error Rate (WER) per-sample maximum (lower is better) | 68.0 % | 43.8 % | Sonic 3.6 slightly ahead |
| Coval TTFA leading silence mean (lower is better) | 193 ms | 135 ms | Sonic 3.6 slightly ahead |
| Coval TTFA roundtrip component mean (lower is better) | 114 ms | 151 ms | Aura-2 slightly ahead |
| Coval pooled corpus Word Error Rate (WER) (lower is better) | 1.9 % | 1.5 % | Sonic 3.6 slightly ahead |
| Coval TTFA roundtrip component p50 (lower is better) | 129 ms | 144 ms | Aura-2 slightly ahead |
| Coval Time to First Audio (TTFA) mean (lower is better) | 307 ms | 286 ms | Sonic 3.6 slightly ahead |
| Coval TTFA leading silence p50 (lower is better) | 134 ms | 137 ms | level |
| Coval Time to First Audio (TTFA) p50 (lower is better) | 279 ms | 279 ms | level |
| Coval Word Error Rate (WER) per-sample p50 (lower is better) | 0.0 % | 0.0 % | level |
Source · latest November 5, 2026
Cekura
Aura-2 slightly aheadAura-2 is clearly ahead in 0 of 9 subtests and slightly in 4, Sonic 3.6 in 0 clearly and slightly in 1. 4 are level.
Cekura TTS stopping duration
lower is better, as of October 19, 2026
Cekura TTS streaming latency median
lower is better, as of October 19, 2026
Cekura TTS hard-part accuracy: read from context
as of October 19, 2026
Show all 9 subtests
| Subtest | Aura-2 | Sonic 3.6 | Verdict |
|---|---|---|---|
| Cekura TTS stopping duration (lower is better) | 40 ms | 764 ms | Aura-2 slightly ahead |
| Cekura TTS streaming latency median (lower is better) | 126 ms | 270 ms | Aura-2 slightly ahead |
| Cekura TTS hard-part accuracy: read from context | 75.0 % | 58.0 % | Aura-2 slightly ahead |
| Cekura TTS hard-part accuracy: medical terms | 86.0 % | 100.0 % | Sonic 3.6 slightly ahead |
| Cekura TTS hard-part accuracy: codes and addresses | 87.0 % | 75.0 % | Aura-2 slightly ahead |
| Cekura TTS ASR-based hard-part accuracy estimate | 77.2 % | 74.9 % | level |
| Cekura TTS non-streaming latency (lower is better) | 281 ms | 286 ms | level |
| Cekura UTMOS-predicted naturalness | 4 | 4 | level |
| Cekura TTS hard-part accuracy: numbers and shorthand | 71.0 % | 71.0 % | level |
Source · latest October 19, 2026
All data side by side
Prices are verbatim from the provider, because audio models bill per minute, character or token.