RealWorldShop - Session Quality: leaderboard
RealWorldShop's session-level process quality: an LLM judge scores clarification, state tracking, constraint handling, pacing and personalization over each full shopping session on a 0-2 scale, averaged; higher is better.
Metric: Session-level process quality (0-2): mean of the GPT-5.5 judge's clarification, state tracking, constraint handling, pacing and personalization scores over each full shopping session; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked. Updated 2 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 1.2 | #177 |
| 2 | Kimi K2.6 | 0.99 | #191 |
| 3 | DeepSeek V3.2 | 0.8 | #417 |
| 4 | Gemini 2.5 Flash | 0.7 | #467 |
| 5 | Qwen 3.5 27B | 0.7 | #374 |
| 6 | Qwen 3.5 35B A3B | 0.64 | #490 |
| 7 | GPT-4o | 0.56 | #754 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=realworldshop-session-quality · How It Works · Data refreshed daily, snapshot 2026-11-17.