Inside LlamaBox 1.1.0
Compare models on your own Android phone
LlamaBox 1.1.0 includes built-in Text, Vision and Docs benchmark tasks for installed models. Results describe your device and configuration.
Choose the task
- Text: writing speed and response latency.
- Vision: picture processing with a compatible model and projector.
- Docs: answering from built-in document material; search work requires the installed search model.
Run a comparison
- Open Settings → Tools → Benchmark.
- Select eligible installed models and tasks.
- Check estimated runtime and memory before starting.
- Allow the run to complete and review its local result history.
A result belongs to a configuration
Record device, Android version, app build, exact model file and quantization, context, threads and thermal conditions. Repeat under comparable conditions if you want to compare changes. The pre-run timing is an estimate; completed run values are measurements.
Speed is one part of the choice
A faster answer does not imply better reasoning or reliable document answers. Built-in task checks do not certify general accuracy. No universal speed ranking or unmeasured performance numbers are published here.