GISA - Set F1: leaderboard
Metric: F1 (%) between the predicted and gold sets on the 50 set-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 February 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 (Thinking) | 64.86 | #138 (Claude Sonnet 4.5) |
| 2 | Gemini 3 Pro (Low) | 63.82 | #77 (Gemini 3 Pro) |
| 3 | Qwen 3 Max (Thinking) | 63.45 | #201 (Qwen 3 Max) |
| 4 | O4 Mini Deep Research | 63.03 | #219 |
| 5 | GPT-5.2 (Thinking) | 62.7 | #105 (GPT-5.2) |
| 6 | Gemini 3 Pro (High) | 62.66 | #77 (Gemini 3 Pro) |
| 7 | Kimi K2.5 (Thinking) | 61.71 | #139 (Kimi K2.5) |
| 8 | Claude Sonnet 4.5 | 60.87 | #138 |
| 9 | DeepSeek V3.2 (Thinking) | 60.79 | #198 (DeepSeek V3.2) |
| 10 | GLM-4.7 (Thinking) | 59.44 | #185 (GLM-4.7) |
| 11 | Qwen 3 235B A22B (Thinking) | 52.37 | #304 (Qwen 3 235B A22B) |
| 12 | DeepSeek V3.2 (Non-reasoning) | 52 | #198 (DeepSeek V3.2) |
| 13 | GPT-4o Search Preview | 38.7 | #445 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=gisa-set-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.