Machine-readable data behind this page: /api/overview?refusals=fallback (JSON). How to query EthIQ: /llms.txt.
Ranked by pass rate. Filters: refusals=fallback. Questions in scope: 376. Models without full coverage of these questions are omitted unless include_partial=true.
| Rank | Model ID | Model | Lab | Effort | Prompt | Pass rate | 95% CI | Questions | Avg cost/question |
|---|---|---|---|---|---|---|---|---|---|
| 1 | gpt-5.3-codex-xhigh | gpt-5.3-codex | openai | xhigh | bare | 81.4% | 77.1% to 85.0% | 376 | $0.1761 |
| 2 | gpt-5.3-codex-high | gpt-5.3-codex | openai | high | bare | 79.5% | 75.2% to 83.3% | 376 | $0.0993 |
| 3 | gpt-5.4-xhigh | gpt-5.4 | openai | xhigh | bare | 76.9% | 72.3% to 80.8% | 376 | $0.1722 |
| 4 | gpt-5.3-codex-medium | gpt-5.3-codex | openai | medium | bare | 70.7% | 66.0% to 75.1% | 376 | $0.0375 |