Trip+ - User Simulation: leaderboard

Metric: Simulated traveler experience (0-100), a rate scaled by 100: median of four profile-conditioned LLM judges (Qwen, Claude, Gemini and GPT families) rating each activity of the final plan for comfort and preference fit, averaged over tasks, on multi-turn personalized travel planning with traveler profiles, request changes and environment disruptions, the same OpenAI-compatible function-calling scaffold over a fixed travel sandbox for every model, temperature 0 where supported; higher is better. Source: arxiv.org. Saturation forecast: Around 2033. 18 models tracked.

Top models

#ModelScore
1GPT-5.451.84
2Qwen 3.6 27B (Non-reasoning)51.67
3Kimi K2.651.64
4Qwen 3.6 35B A3B (Non-reasoning)51.41
5Seed 2.0 Pro51.28
6DeepSeek V4 Pro51
7GPT-5.4 Mini50.85
8Gemini 3.1 Pro (Preview)50.77
9GLM-5.150.14
10MiniMax-M2.749.92
11Gemma 4 31B49.23
12Gemma 4 26B A4B49.16
13Hy3-preview48.66
14Gemini 3 Flash (Preview)48.62
15Qwen 3.5 27B (Non-reasoning)48.21

Interactive version: theaggregate.ai/benchmark?slug=trip-plus-user-simulation · How It Works · Data refreshed daily, snapshot 2026-09-29.