E-commerce Last Exam: leaderboard
FlyaiLab’s agentic shopping and travel benchmark: the agent has to complete a multi-step purchase or booking against a simulated storefront, and the score is mean reward over the episode rather than a pass or fail. This board is the overall configuration; the travel and e-commerce boards split it.
Metric: Mean Reward (%). Source: huggingface.co. Saturation forecast: Around 2032. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 | 56.74 | #30 |
| 2 | Grok 4.5 | 56.31 | #48 |
| 3 | Claude Opus 5 | 54.35 | #12 |
| 4 | Claude Opus 4.8 | 53.97 | #37 |
| 5 | Claude Opus 4.7 | 53.5 | #49 |
| 6 | GPT-5.6 Sol | 53.17 | #19 |
| 7 | Gemini 3.7 Flash | 53.17 | #23 |
| 8 | GPT-5.5 | 53.09 | #31 |
| 9 | Qwen 3.8 Max | 53.04 | #35 |
| 10 | Claude Sonnet 5 | 53.03 | #57 |
| 11 | Claude Opus 4.6 | 52.92 | #63 |
| 12 | GLM-5.2 | 52.82 | #73 |
| 13 | Gemini 3.6 Flash | 52.59 | #51 |
| 14 | Qwen 3.8 Flash | 52.59 | #50 |
| 15 | MiMo-V2.5-Pro | 52.57 | #158 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=e-commerce-last-exam · How It Works · Data refreshed daily, snapshot 2026-10-20.