ConsumerSimBench: leaderboard
LLMs are increasingly used as "digital consumers" to simulate public opinion, pre-test marketing decisions, and anticipate audience response. However, existing evaluations rarely ask whether a model can reconstruct the concrete reaction patterns that real consumers surface in public discourse. We introduce ConsumerSimBench, a benchmark built from 1,553 real Chinese social-media topics and 23,122
Metric: Avg (95% CI) (self-reported). Source: benchmarklist.com. Saturation forecast: Around September 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 47.8 | #58 |
| 2 | Gemini 3 Flash | 46.6 | #99 |
| 3 | Grok 4.20 | 44.9 | #175 |
| 4 | Kimi K2.5 | 43.5 | #144 |
| 5 | Qwen 3.5 397B A17B | 43.4 | #146 |
| 6 | DeepSeek V3 | 42.9 | #321 |
| 7 | Seed 1.8 | 42.4 | #139 |
| 8 | GPT-5.2 | 35.8 | #110 |
| 9 | MiniMax-M2.5 | 35.2 | #303 |
| 10 | Claude Opus 4.6 | 32.8 | #63 |
| 11 | MiMo-V2.5-Pro | 32.3 | #158 |
| 12 | Claude Sonnet 4.6 | 30.8 | #90 |
| 13 | GPT-4o | 29.7 | #343 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=consumersimbench · How It Works · Data refreshed daily, snapshot 2026-10-20.