Benchmarks
In-browser performance depends on your GPU, your CPU and your browser, so this page does not publish a leaderboard of numbers we collected on our machines. Instead: what the metrics mean, how they are measured, and a curated table of published measurements from external sources — each attributed. For numbers on your hardware, run the live demos on the model pages.
What is measured, and how
Real-time factor (TTS)
Tokens per second (LLM)
Word error rate (STT)
STT accuracy references we link to per model: the Open ASR Leaderboard and each model's Hugging Face model card, linked from every model page.
Published measurements (external)
These are external, published measurements — not run by TTSLab, and not reproducible on your machine without its exact context. Each row names its source. To see what a model does on your hardware, run its live demo.
| Model / system | Metric | Published result | Context | Source |
|---|---|---|---|---|
| Kokoro-82M — WebGPU backend | Real-time factor (RTF) | 0.105 — about 9.5x faster than real time | Apple M1 Max, in-browser | briantung.me, Aug 2026 |
| Kokoro-82M — single-thread WASM (q8) | Real-time factor (RTF) | 2.002 — about half of real time | Apple M1 Max, in-browser (same machine as the WebGPU row) | briantung.me, Aug 2026 |
| Llama-3.1-8B via WebLLM | Decode speed | 41.1 tokens/sec — roughly 71% of native Metal performance | Apple M3 Max, in-browser | WebLLM paper (arXiv 2412.15803) |
| Whisper tiny / base via whisper.cpp | Transcription speed | Roughly 2-3x faster than real time on modern CPUs | WASM / CPU, in-browser | whisper.cpp README |
| Moonshine | CPU transcription latency | Paper claims 5-15x faster than comparably accurate Whisper models on CPU | CPU, on-device | Moonshine paper (arXiv 2410.15608) |
Planned: a standardized suite
A reproducible in-browser benchmark suite — fixed test phrases for TTS, fixed clips for STT, results anchored to your device profile — is planned. Behind the scenes, generations on this site already pass through an automated QA harness that checks each render for truncation, looping, clipping and silence before it counts as a success. Subscribe below if you want to know when the public leaderboard lands.