Gradium says Phonon on-device TTS passed 200,000 downloads in two weeks
Gradium reported on November 5, 2026 that its Phonon on-device text-to-speech models have been downloaded more than 200,000 times in their first two weeks. The company says Phonon ranked first for intelligibility in German and Spanish on the Open TTS Leaderboard.
Gradium reported on November 5, 2026 that its Phonon on-device text-to-speech (TTS) models have been downloaded more than 200,000 times in their first two weeks. The company published the figure in a technical follow-up on Hugging Face, as covered by Unite.AI.
Benchmark claims
According to Gradium, Phonon ranked first for intelligibility in German and Spanish on the Open TTS Leaderboard. The company also says Phonon placed in the top three in every language it evaluated, alongside models up to 100 times its size.
Gradium also published a chart of 95th-percentile CPU time to first audio, which placed Phonon 31m English first. These results are Gradium's own reporting.
Key figures
- Downloads: more than 200,000 in the first two weeks, per Gradium
- Intelligibility: first in German and Spanish on the Open TTS Leaderboard, per Gradium
- Overall ranking: top three in every language evaluated, per Gradium
- Comparison set: models up to 100 times the size of Phonon
- Latency metric: 95th-percentile CPU time to first audio, where Phonon 31m English ranked first on Gradium's chart
What this means
The download count shows early uptake of Phonon as an on-device TTS option. The leaderboard and latency results, as Gradium presents them, point to small models that compete on intelligibility with much larger ones while running on CPU. Because the rankings and the latency chart come from Gradium's own technical follow-up, readers evaluating Phonon for deployment should check the Open TTS Leaderboard and run their own tests on target hardware.
Related Articles
Doubao Work tops SuperCLUE-Office October benchmark at 85.33, ahead of Kimi Work (84.57)
ByteDance's Doubao Work took first place in SuperCLUE's October 2026 SuperCLUE-Office benchmark with 85.33, ahead of Moonshot AI's Kimi Work (84.57) and Tencent's WorkBuddy (83.99). Eight Chinese office agents were scored on Word, PowerPoint, Excel and chart deliverables.
CloudOpsBench: Claude Fable 5.1 solves 60% of AWS DevOps tasks, GPT-6 Astra 53%
CloudOpsBench is a new open-source benchmark of 45 AWS DevOps tasks. Claude Fable 5.1 solved 60 percent and GPT-6 Astra 53 percent, and no model completed every task. The research team says reasoning, not knowledge, is the main bottleneck.
NVIDIA lifts Ling-3.0-flash GSM8K decode throughput 42.5% on a single DGX Spark with 3 MoE optimizations
NVIDIA describes 3 stacked optimizations for Mixture-of-Experts (MoE) inference on a single DGX Spark. On Ant Group inclusionAI's 124B Ling-3.0-flash, GSM8K decode throughput rose 42.5% with no detectable accuracy change.
Google releases 29 public tasks from 85-task engineering benchmark PLE and opens agent competition
Google AI & Infrastructure announced on November 3, 2026 that it is releasing 29 public tasks, grading tools and an evaluation container for Physical-Engineering's Last Exam (PLE). The 85-task benchmark was developed with Mercor and comes with an open agent competition and leaderboard.
Comments
Loading...