Today for AI
HOT RADAR
agentHEAT 11.3°

AI CURATED HOT STORY · 10/26/2026

Google Releases Android Bench 2.0 with Long-Horizon Tasks and Continuous Scoring

1 reports archived1 independent sourcesupdated 10/26/2026, 11:33:21
Synthesis & Latest Updates
1 Sources Cross-Validated

Google released Android Bench 2.0, introducing Long-Horizon Tasks (LHT), agent-based evaluation, and continuous scoring to better assess complex multi-step development tasks. The benchmark reveals that AI performs better at writing new code than refactoring existing code, with Claude Opus 5.5 currently leading the leaderboard with a 32% pass rate.

LATEST/Google released Android Bench 2.0, introducing Long-Horizon Tasks (LHT), agent-based evaluation, and continuous scoring to better assess complex multi-step development tasks. The benchmark reveals that AI performs better at writing new code than refactoring existing code, with Claude Opus 5.5 currently leading the leaderboard with a 32% pass rate.

Hourly heat curve

3 fully observed hours
Now
11.3
Peak
11.912:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 1 reports · Latest first
  1. InfoQ 中文T2·78 pts

    Google Releases Android Bench 2.0 with Long-Horizon Tasks and Continuous Scoring

    • Introduces Long-Horizon Tasks (LHT) covering practices like dependency upgrades and zero-to-one builds that take days to complete
    • Adopts continuous scoring instead of binary pass/fail to improve recognition of partial successes
    • Data shows AI excels more at writing new code than refactoring; Claude Opus 5.5 currently tops the leaderboard