T2S-Bench - Link Extraction: leaderboard
Metric: Link F1 (%) on T2S-Bench-E2E (87 human-verified text-structure pairs): given the text and every node, the model outputs the links as JSON, scored by F1 against the reference links, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 45 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 86.91 | #138 |
| 2 | Gemini 2.5 Pro | 84.32 | #145 |
| 3 | Claude Sonnet 4 (20250514) | 84.07 | #211 |
| 4 | DeepSeek R1 0528 | 80.31 | #217 |
| 5 | GPT-5.1 | 79.44 | #131 |
| 6 | Claude Haiku 4.5 (20251001) | 79.33 | #251 |
| 7 | DeepSeek V3.2 | 78.69 | #198 |
| 8 | DeepSeek V3 (0324) | 78.17 | #332 |
| 9 | DeepSeek V3.1 | 77.77 | #260 |
| 10 | GPT-5.2 | 77.76 | #105 |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 77.64 | #306 (Qwen 3 Next 80B A3B) |
| 12 | Qwen 3 Next 80B A3B Instruct | 76.11 | #352 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 76.11 | #253 (Qwen 3 235B A22B 2507) |
| 14 | Claude 3 Haiku (20240307) | 75.51 | #759 |
| 15 | Gemini 2.5 Flash | 75.1 | #237 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=t2s-bench-link-extraction · How It Works · Data refreshed daily, snapshot 2026-10-11.