Comparison · read on October 27, 2026
Gradium TTS vs MiniMax Speech 2.8 HD
Two text-to-speech (TTS) models side by side, from four public benchmarks. Speed comes first, then listener rating and price; the three are never added into a score.
| Measure | Gradium TTS | MiniMax Speech 2.8 HD |
|---|---|---|
| Provider | Gradium | MiniMax |
| Speed tier | Headroom | Not timed |
| P95 time to first audio | 104.3 ms | Not timed |
| Median time to first audio | 53.7 ms | Not timed |
| Listener Elo | 1,174 | 1,174 |
| Price per 1M characters | $36 to $57.8 | $100 |
| Access | Proprietary | Proprietary |
| Rank in the full ranking | 2 | 42 |
| Rank by use case | Customer service 1, Voice agents 2, Assistants 1, Entertainment 1, Knowledge sharing 1 | Customer service 29, Assistants 29, Entertainment 29, Knowledge sharing 29 |
Which should you choose?
- Speed: not compared, because a benchmark in this index does not publish it for both models.
- Listener rating: Gradium TTS and MiniMax Speech 2.8 HD are a tie (Elo 1,174 against 1,174; scores within each other's 95% intervals are a tie).
- Price: Gradium TTS comes out ahead ($36 to $57.8 against $100 per million characters, lowest list price).
Where a model is not timed or not rated, nothing is estimated. See each model page for its sources: Gradium TTS and MiniMax Speech 2.8 HD.
Frequently asked questions
- Which is faster, Gradium TTS or MiniMax Speech 2.8 HD?
- Speed: not compared, because a benchmark in this index does not publish it for both models.
- Which is rated higher by listeners, Gradium TTS or MiniMax Speech 2.8 HD?
- Listener rating: Gradium TTS and MiniMax Speech 2.8 HD are a tie (Elo 1,174 against 1,174; scores within each other's 95% intervals are a tie).
- Which is cheaper, Gradium TTS or MiniMax Speech 2.8 HD?
- Price: Gradium TTS comes out ahead ($36 to $57.8 against $100 per million characters, lowest list price).
- Why is this model always on the left?
- Gradium TTS is the reference model of this site's comparison pages: it is set against every other listed model. The editor chose it; it is not chosen by rank.