TTS API comparison 2026
Gradium TTS vs OpenAI GPT-Realtime-2 Gradium vs OpenAI · text-to-speech for voice agents
Gradium TTS is in the Headroom speed tier (P95 at or under 101 ms); OpenAI GPT-Realtime-2 is in the no speed tier (not measured by Coval). 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 | Gradium TTS | OpenAI GPT-Realtime-2 |
|---|---|---|
| Speed tierCoval P95, 30 days | Headroom | not measured |
| Median time to first audioCoval, 30-day window | 55 ms | — |
| P95 time to first audio (slow turns)Coval, 30-day window | 101 ms | — |
| Time to first audio, board chartSpeko, US East | 55 ms | — |
| Naturalness EloSpeko | 1,592 | not rated |
| Listener Elo (95% interval)Artificial Analysis | 1,176 ±13 | 1,075 ±12 |
| Numbers and dates (0 to 1)Speko | 0.27 | not rated |
| List price per 1M charactersCoval, Speko, Artificial Analysis | $35.9 to $57.8 | $191.6 |
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
- →Listener Elo (95% interval): 1,176 ±13 vs 1,075 ±12 (Artificial Analysis)
- →List price per 1M characters: $35.9 to $57.8 vs $191.6 (Coval, Speko, Artificial Analysis)
Where OpenAI GPT-Realtime-2 is ahead
On the published measures, OpenAI GPT-Realtime-2 is ahead of Gradium TTS on none beyond the tie rule. This says nothing about features the benchmarks do not measure.
OpenAI GPT-Realtime-2: all measurements →Frequently asked questions
- What is the difference between Gradium TTS and OpenAI GPT-Realtime-2?
- Gradium TTS (Gradium) is in the Headroom speed tier (P95 at or under 101 ms); OpenAI GPT-Realtime-2 (OpenAI) is in the no speed tier (not measured by Coval). Gradium TTS leads on listener Elo (95% interval) and list price per 1M characters; OpenAI GPT-Realtime-2 leads on no measure. No combined score is computed.
- Which is faster for real-time voice agents, Gradium TTS or OpenAI GPT-Realtime-2?
- Only Gradium TTS is measured by Coval (median 55 ms, P95 101 ms); OpenAI GPT-Realtime-2 has no speed tier, so the two cannot be compared on the same window.
- Which sounds more natural, Gradium TTS or OpenAI GPT-Realtime-2?
- Speko naturalness: 1,592 for Gradium TTS, not rated for OpenAI GPT-Realtime-2. Artificial Analysis listener Elo: 1,176 ±13 and 1,075 ±12. 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.