ComboShoppingBench - Budget Compliance: leaderboard
Metric: Pass rate (%). Source: arxiv.org. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (Thinking) | 87.3 |
| 2 | GPT-5.5 | 86.3 |
| 3 | Claude Opus 4.6 (Non-reasoning) | 82.5 |
| 4 | Kimi K2.6 (Thinking) | 80.8 |
| 5 | Claude Opus 4.6 (Thinking) | 80.4 |
| 6 | Claude Opus 4.8 (Thinking) | 80.1 |
| 7 | GLM-5.2 (Non-reasoning) | 78.4 |
| 8 | Claude Sonnet 4.6 (Non-reasoning) | 78 |
| 9 | Claude Sonnet 4.6 (Thinking) | 74.9 |
| 10 | Qwen 3.6 27B (Non-reasoning) | 74.6 |
| 11 | Seed 2.0 Pro (Thinking) | 72.9 |
| 12 | DeepSeek V4 Pro (Non-reasoning) | 72.5 |
| 13 | MiniMax M3 (Thinking) | 66.3 |
| 14 | Kimi K2.6 (Non-reasoning) | 65.6 |
| 15 | Qwen 3.6 27B (Thinking) | 63.2 |
Interactive version: theaggregate.ai/benchmark?slug=comboshoppingbench-budget-compliance · How It Works · Data refreshed daily, snapshot 2026-09-19.