Know what your phone can actually run

EdgeFit AI measures real language-model inference on physical Android phones — then proves which results are comparable. A prototype built for the Meta Global AI Developer Hackathon.

See evidence status How it works
EdgeFitDiag Home screen on a Samsung phone: device overview with CPU, RAM, battery and thermal metric cards

Captured on-device real screenshot — status bar and time are from the original capture.

Hardware-aware, like the app

The Android app starts from your device, not from a leaderboard. The same idea, on the web:

Example device profile DEMONSTRATION

Illustrative values for a phone of this class — not measured from your browser, which cannot see your hardware.

Device
SM-S908U class
CPU cores
8
Total RAM
≈7.1 GB visible (example)
Free memory
~1.8 GB (illustrative, varies live)
Thermal
nominal (illustrative)
ABI
arm64-v8a

Try the model-fit check DEMONSTRATION

Pick a model size to see how EdgeFit reasons about fit before any benchmark runs. Sizes are public model facts; memory figures are illustrative. This is arithmetic, not a measurement — EdgeFit shows throughput only from real qualified runs.

The Android workflow

  1. 1Select a GGUF model file on the phone.
  2. 2Configure inference: context, threads, prediction length.
  3. 3Run a physical benchmark; results stay on-device.
  4. 4Review measured throughput per session.
  5. 5Optionally contribute with explicit consent (in development; staging backend under test, not publicly queryable).

App sections, as shipped: Home · Benchmark · Sessions · Agent · Settings.

EdgeFitDiag Benchmark screen: numbered Model, Configure and Run sections with model file and config actions

Captured on-device real screenshot — status bar and time are from the original capture.

Meta Muse Spark, honestly scoped

EdgeFit uses Meta's Muse Spark model to help analyze validated evidence: a validated tool request runs one fixed read-only query over local SQLite, and the model reasons over the returned rows. No live model outputs are shown on this site, and nothing here implies endorsement by Meta.

Global evidence, gated by proof

The long-term goal is a global network where only verified, consented, comparable phone measurements count. Today: zero verified global aggregates published. The evidence status page tracks exactly what exists and what publication requires.

Provenance honesty

All EdgeFit AI source in this repository is new work started October 26, 2026, for this hackathon. Honey LLM and BeeBench are pre-existing independent projects; EdgeFit does not reuse their code or present their results as its own.