TTS API comparison 2026
Gradium TTS vs SpaceXAI Grok TTS Gradium vs xAI · text-to-speech for voice agents
Gradium TTS is in the Headroom speed tier (P95 under 120 ms); SpaceXAI Grok TTS is in the Visible delay speed tier (P95 over 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 | SpaceXAI Grok TTS |
|---|---|---|
| Speed tierCoval P95, 30 days | Headroom | Visible delay |
| Median time to first audioCoval, 30-day window | 54 ms | 330 ms |
| P95 time to first audio (slow turns)Coval, 30-day window | 104 ms | 474 ms |
| Time to first audio, board chartSpeko, US East | 55 ms | 258 ms |
| Time to first audio, full sentence (median)Cekura | 76 ms | 358 ms |
| Time to first audio, text streamed word by word (median)Cekura | -64 ms | 157 ms |
| Naturalness EloSpeko | 1,592 | 1,495 |
| Listener Elo (95% interval)Artificial Analysis | 1,174 ±13tie | 1,155 ±13tie |
| Numbers and dates (0 to 1)Speko | 0.27 | 0.45 |
| Hard passages pronounced correctly (95% interval)Artificial Analysis, human reviewers | not tested | 87.6% (85.2 to 89.9) |
| Reading accuracy on hard textCekura, transcriber-scored | 79.5% | 87.4% |
| List price per 1M charactersCoval, Speko, Artificial Analysis, Cekura | $36 to $57.8 | $15 |
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 330 ms (Coval, 30-day window)
- →P95 time to first audio (slow turns): 104 ms vs 474 ms (Coval, 30-day window)
- →Time to first audio, board chart: 55 ms vs 258 ms (Speko, US East)
- →Time to first audio, full sentence (median): 76 ms vs 358 ms (Cekura)
- →Time to first audio, text streamed word by word (median): -64 ms vs 157 ms (Cekura)
- →Naturalness Elo: 1,592 vs 1,495 (Speko)
Where SpaceXAI Grok TTS is ahead
- →Numbers and dates (0 to 1): 0.45 vs 0.27 (Speko)
- →Reading accuracy on hard text: 87.4% vs 79.5% (Cekura, transcriber-scored)
- →List price per 1M characters: $15 vs $36 to $57.8 (Coval, Speko, Artificial Analysis, Cekura)
Frequently asked questions
- What is the difference between Gradium TTS and SpaceXAI Grok TTS?
- Gradium TTS (Gradium) is in the Headroom speed tier (P95 under 120 ms); SpaceXAI Grok TTS (xAI) is in the Visible delay speed tier (P95 over 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 naturalness Elo; SpaceXAI Grok TTS leads on numbers and dates (0 to 1), reading accuracy on hard text and list price per 1M characters. They are a tie on listener Elo (95% interval). No combined score is computed.
- Which is faster for real-time voice agents, Gradium TTS or SpaceXAI Grok TTS?
- On Coval's 30-day window as of October 27, 2026, the median time to first audio is 54 ms for Gradium TTS and 330 ms for SpaceXAI Grok TTS (Gradium TTS ahead by 276 ms); the P95, the slow turn, is 104 ms and 474 ms (Gradium TTS ahead by 370 ms). Gaps under 10 ms are ties.
- Which sounds more natural, Gradium TTS or SpaceXAI Grok TTS?
- Speko naturalness: 1,592 for Gradium TTS, 1,495 for SpaceXAI Grok TTS. Artificial Analysis listener Elo: 1,174 ±13 and 1,155 ±13, a tie because the 95% intervals overlap. 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.