IBARA LAB
Benchmark archive
Versioned computer-use and browser-use benchmark reports.
Updated
Each report preserves its measurement dates, method, candidate, outcome counts and limitations. Campaigns with different tasks or configurations are not pooled into a reliability score.
- Permanent Ubuntu peer and reader calibration journal: revision 1: Diagnostic run · 86 attempts recorded. Measured 2026-10-11; benchmark.
- Ubuntu common native confirmation: revision 1: Diagnostic run · 60 attempts recorded. Measured 2026-10-11; benchmark.
- Omarchy common native confirmation: revision 1: Diagnostic run · 60 attempts recorded. Measured 2026-10-11; benchmark.
- Ubuntu interrupted common native diagnostic: revision 1: Diagnostic run · 40 attempts recorded. Measured 2026-10-11; benchmark.
- Omarchy interrupted common native diagnostic: revision 1: Diagnostic run · 41 attempts recorded. Measured 2026-10-11; benchmark.
- Ubuntu native benchmark confirmation: revision 1: Diagnostic run · 96 attempts recorded. Measured 2026-10-11; benchmark.
- Omarchy native benchmark confirmation: revision 1: Diagnostic run · 96 attempts recorded. Measured 2026-10-10; benchmark.
- Native interaction benchmark, round 4: revision 2: Diagnostic run · 60 attempts recorded. Measured 2026-10-10; benchmark.
- Six workflows with Sonnet 5.5 and GPT-6.1 Sol at medium effort: revision 1: Diagnostic run · 12 attempts recorded. Measured 2026-10-09; benchmark.
- Six-task computer-use baseline with GPT-6.1 Sol: revision 1: Diagnostic run · 6 attempts recorded. Measured 2026-10-09; benchmark.
- Incumbent computer-use diagnostic campaign: revision 1: Incomplete benchmark · 65 attempts recorded. Measured 2026-10-08; benchmark.
- Computer and browser use: diagnostic pilot: revision 3: Diagnostic run · 22 attempts recorded. Measured 2026-10-08; benchmark.
- Computer and browser use: ibara 0.1.0-17: revision 1: 132/157 passed (84.1%). Measured 2026-09-29; historical.