GISA - Table Item F1: leaderboard
Metric: Item-level F1 (%) between predicted and gold table cells on the 253 table-type queries; GISA's human-written information-seeking queries with deterministic answers (normalized TSV output compared cell by cell); LLMs run as ReAct agents with Google search (Serper) and Jina browse tools whose pages the same model summarizes, at most 30 tool calls and 8,192 output tokens per step, while commercial search systems run as offered; higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Max (Thinking) | 66.86 | #201 (Qwen 3 Max) |
| 2 | Gemini 3 Pro (High) | 66.02 | #77 (Gemini 3 Pro) |
| 3 | Claude Sonnet 4.5 (Thinking) | 65.17 | #138 (Claude Sonnet 4.5) |
| 4 | Gemini 3 Pro (Low) | 64.93 | #77 (Gemini 3 Pro) |
| 5 | Claude Sonnet 4.5 | 63.71 | #138 |
| 6 | DeepSeek V3.2 (Thinking) | 62.42 | #198 (DeepSeek V3.2) |
| 7 | DeepSeek V3.2 (Non-reasoning) | 62.24 | #198 (DeepSeek V3.2) |
| 8 | GLM-4.7 (Thinking) | 61.28 | #185 (GLM-4.7) |
| 9 | Kimi K2.5 (Thinking) | 61.23 | #139 (Kimi K2.5) |
| 10 | GPT-5.2 (Thinking) | 60.2 | #105 (GPT-5.2) |
| 11 | O4 Mini Deep Research | 56.47 | #219 |
| 12 | GPT-4o Search Preview | 45.61 | #445 |
| 13 | Qwen 3 235B A22B (Thinking) | 43.93 | #304 (Qwen 3 235B A22B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=gisa-table-item-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.