Source Attribution on 130 research queries (test)
100Success RateGPT-5.4
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-5.42026.05 | 100 | 100 | 93.7 | 47.7 | |
| GPT-5.22026.05 | 100 | 98.3 | 92.3 | 58.8 | |
| Codex2026.05 | 100 | 96.9 | 91.9 | 54.1 | |
| GPT-5 Mini2026.05 | 100 | 99.3 | 87.4 | 38.9 | |
| Gemini 3 Flash2026.05 | 100 | 94.7 | 82.9 | 45.2 | |
| Claude Sonnet 4.52026.05 | 96.7 | 98.9 | 88.3 | 51.8 | |
| Claude Sonnet 4.62026.05 | 93.3 | 99.2 | 89.8 | 58.7 | |
| Claude Opus 4.62026.05 | 93.3 | 97.2 | 83.9 | 54.2 | |
| Claude Opus 4.52026.05 | 90 | 98.7 | 95.7 | 76.8 | |
| Gemini 3.1 Pro2026.05 | 90 | 94.1 | 80.7 | 48.5 | |
| Claude Haiku 4.52026.05 | 83.3 | 98.9 | 91.1 | 68.9 | |
| OSS-120B2026.05 | 40 | 83.9 | 68.7 | 24.4 | |
| Llama 4 Maverick2026.05 | 30 | 80.8 | 60.6 | 34.3 | |
| Pixtral Large2026.05 | 16.7 | 100 | 64.9 | 51.4 |