Detected locally
Checking this browser…
- WebAssembly
- Checking
- WebGPU
- Checking
- CPU threads
- Checking
- Browser storage
- Browser managed
- Reported memory
- Not exposed
The profile stays in this tab and is not sent to a server.
Product 02 · Device & Model Matcher
Choose a model your device can actually run.
Inspect this browser, run an optional local benchmark, then match real browser-ready models by runtime, download size, hardware, and task.
StatusLiveEvery Run Any AI product declares what is real today.
Your device profile
Match before downloading.
These are capability-based recommendations, not guarantees. A real model run is the final compatibility test.
Not run
Run a short arithmetic workload to add a rough CPU signal. It downloads no model and does not measure GPU performance.
No cached models
Inspecting browser model cache…
Matched models
Three models you can test now.
Compatible
Classification · q8
DistilBERT SST-2 · quantized
Classifies short English text as positive or negative using the browser CPU.
Checking this browser≈70 MB first download · CPU
- Runtime
- WebAssembly
- Language
- English
- License
- Apache-2.0
Compatible
Embeddings · q8
all-MiniLM-L6-v2 · quantized
Turns two English passages into normalized 384-dimensional vectors and compares their semantic similarity.
Checking this browser≈25 MB first download · CPU
- Runtime
- WebAssembly
- Language
- English
- License
- Apache-2.0
Compatible
Transcription · q8 CPU · fp32/q4 WebGPU
Whisper Tiny English · quantized
Transcribes a short English audio file locally using WebGPU when available or a slower CPU fallback.
Checking this browser≈45–125 MB first download · CPU or WebGPU
- Runtime
- WebGPU with WebAssembly fallback
- Language
- English
- License
- Apache-2.0
Recommendations use browser capability signals, storage estimates, and conservative runtime rules. Actual speed varies by browser, operating system, drivers, thermal limits, input length, and other open applications.