Agentic Reliability on HOB
7.75Task Success (TS)GPT-4.1
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-4.1Agent backbone=GPT-4.12026.06 | 7.75 | 6.4 | 6.1 | 6.75 | 6.2 | |
| Claude Opus 4.8Agent backbone=Claude Opus 4.82026.06 | 7.2 | 8.9 | 8.9 | 8.6 | 8.7 | |
| GPT-5.5Agent backbone=GPT-5.52026.06 | 6 | 7.07 | 6.53 | 7.33 | 6.33 | |
| Claude Opus 4.7Agent backbone=Claude Opus 4.72026.06 | 5.75 | 7.75 | 7.88 | 7.27 | 8.31 |