Artificial Analysis launches TTS leaderboard with new Pronunciation Robustness Benchmark across 95 models
ArtificialAnlys · x · 2026-09-24
Artificial Analysis launched a Text-to-Speech comparison page covering 27 of 95 models, comparing quality, speed and price of provider-native voices, ranked by Arena Elo from user votes in its Speech Arena.
It also introduced a new Pronunciation Robustness Benchmark: human reviewers judge the share of English spans pronounced correctly, spanning four categories —
- Standalone terms: brand, place and technical names (e.g. Arkansas, façade);
- Contextual disambiguation: same spelling, different reading by context;
- Plus shorthand expansion and other categories.
Each model is tested with a single voice; prompts are sent as written with no normalization, using the model's default text normalization setting when offered.
More from Models
- ChatGPT Voice gets plugins and GPT-6 power, can now build docs and decks by voice — Dimillian · 2026-09-24
- Leak Chatter: Frontier Model 8 Months Ahead of Astra, Only 2 Months Ahead of Trend — scaling01 · 2026-09-24
- Navier-Stokes proof burned 130B output and trillions of input tokens, dwarfing top OpenAI users' few billion daily — scaling01 · 2026-09-24
- Meta Muse scores 4.88 across 40k ratings — analyst says product, distribution, price make it near-unbeatable — RihardJarc · 2026-09-24
- Audi 3D Modeling Benchmark Rerun: Only GPT-6 Astra and Opus 5.5 Are Worth Taking Seriously — kevinkern · 2026-09-24
- DIY calibration test: 200 items, score accuracy vs self-claimed 90% confidence — colinmcnamara · 2026-09-24