FutureX - Level 3 (Deep Search): leaderboard
FutureX Level 3 (Deep Search tier): open-ended questions about future events whose underlying facts are relatively stable, answered by LLM search agents without options. Official runs from the newest week that has them, so the event set changes weekly. Higher is better, 0-100 (%); weighted 30% overall.
Metric: Level 3 score (latest official week, %). Source: futurex.live. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (High) | 67.54 | #33 (GPT-5.5) |
| 2 | Qwen 3.7 Max | 57.3 | #85 |
| 3 | GPT-5.4 (High) | 44.41 | #83 (GPT-5.4) |
| 4 | Seed 2.0 Pro | 44 | #122 |
| 5 | DeepSeek V3.2 (Thinking) | 35.52 | #210 (DeepSeek V3.2) |
| 6 | Grok 4.20 (Reasoning) | 35.39 | #179 (Grok 4.20) |
| 7 | Seed 2.0 Lite | 33.78 | #161 |
| 8 | Gemini 3.1 Pro (Preview) (High) | 31.94 | #64 (Gemini 3.1 Pro (Preview)) |
| 9 | Qwen 3.6 Max Preview | 27.02 | #91 |
| 10 | MiniMax-M2.7 | 24 | #270 |
| 11 | DeepSeek V4 Flash | 20.44 | #155 |
| 12 | Gemini 3.5 Flash | 17.33 | #66 |
| 13 | MiniMax-M3 | 17.31 | #151 |
| 14 | Grok 4 | 8.44 | #189 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=futurex-level-3-deep-search · How It Works · Data refreshed daily, snapshot 2026-11-05.