AlpsBench - Preference Following (Interactive): leaderboard
Metric: Pass rate (0-1, times 100) of responses to user queries that must follow implicit preferences in interactive requests, judged pass or fail by a DeepSeek-V3.2 judge given the dialogue history; AlpsBench's 2,500 human-verified instances per task built from long-term real user-LLM dialogues in WildChat; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 76.8 | #105 |
| 2 | Gemini 3 Flash (Preview) | 70.52 | #78 |
| 3 | Qwen 3 Max | 65.74 | #201 |
| 4 | DeepSeek V3.2 (Thinking) | 61.2 | #198 (DeepSeek V3.2) |
| 5 | Claude Sonnet 4.5 | 54.98 | #138 |
| 6 | GPT-4.1 Mini | 52.4 | #346 |
| 7 | Llama 4 Maverick | 30.8 | #451 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=alpsbench-preference-following-interactive · How It Works · Data refreshed daily, snapshot 2026-10-11.