AI News Feed
Market watch
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
Products & Applications

Gradium AI Releases Default TTS Model with 81% Hard-Case Pass Rate and 216 ms Latency

Gradium AI's new default TTS model reports 81% hard-case pass rate, beating competitors like Cartesia and ElevenLabs, with 216 ms time-to-first-audio and no migration needed.

AI News Feed · 2026-09-01 13:16

The evaluation set, which the company open-sourced on Hugging Face under CC BY 4.0, consists of 100 items across 10 criteria in five languages: English, German, French, Spanish, and Portuguese. The criteria include spelling, acronyms, alphanumeric tokens, dates, numbers, and email addresses, plus three composite criteria simulating realistic agent turns for orders, IT tickets, and claims. Scoring was conducted by independent native-speaker raters, with a sentence passing only if every element is pronounced correctly and completely.

In the comparison, Gradium TTS scored 81.0%, ahead of Cartesia Sonic 3.6 at 75.1%, ElevenLabs v3 Conversational at 65.4%, Fish Audio S2.1 Pro at 49.5%, and Inworld TTS 1.5 Max at 46.5%. All models were generated in August 2026 with default settings, and audio was loudness-normalized and randomized for the human evaluation.

On latency, Gradium reports a 216 ms P50 time-to-first-audio on Coval, which is 170 ms faster than its previous model. The p75-p25 interquartile range was 30 ms across 480 runs, described as the tightest among the five models tested. By comparison, Cartesia Sonic 3.6 had a 454 ms median and a 165 ms spread, while Inworld TTS 2 posted a 166 ms median, Fish Audio S2.1 Pro a 291 ms median, and ElevenLabs v3 Conversational a 329 ms median. Gradium's claim is about the combination of low failure rate and fast first audio with low variance.

The model is already deployable with no migration required. Existing voices, including custom clones, continue to work unchanged. New teams can install the Python SDK and reuse existing voice IDs. Gradium is also offering 1 million credits for complete hard-case failure reports submitted on its Discord. MarkTechPost notes that the benchmark is vendor-run, but the evaluation set is open under CC BY 4.0.