v148 · on-device AI

Quality Control Panel

Run all three performance modes — fast, balanced, quality — on the same text simultaneously. Compare latency, word-count reduction, and output quality side by side to decide which mode is right for each call site.

—Fastest mode (ms)

—Best compression %

—Inference mode

Input

Sample text

fast

Not run

balanced

Not run

quality

Not run

// Run all three modes in parallel
const MODES = ['fast', 'balanced', 'quality'];

const [fastResult, balancedResult, qualityResult] = await Promise.all(
  MODES.map(async pref => {
    const t0 = performance.now();
    const s = await Summarizer.create({ performancePreference: pref, type: 'tl;dr' });
    const text = await s.summarize(inputText);
    return { pref, text, latency: performance.now() - t0 };
  })
);

implementation reference

Need the exact API surface, compatibility boundaries, errors, lifecycle, and source links? Read the matching gendn reference ↗

see also

implementation reference