Cantonese Voice Benchmark 2026.1 · Window 2
Every published number carries a bootstrap confidence interval. Entries that overlap are reported as tied rather than ordered — with 30 clips a fraction of a percent is noise.
The first scored window has not closed. Reference baselines are shown below so the scale is meaningful the moment real entries land.
| A | Open-source pipeline | 76.2 | 11.1% | 3.81 | 3.44 | 3.29 | 2.11sp95 3.88 |
| B | Proprietary voice LLM | 82.7 | 8.6% | 4.14 | 3.62 | 4.01 | 0.88sp95 1.52 |