S2SServiceBench (Signal Comprehension): leaderboard
Metric: Level 1 overall score (0-1 printed, shown times 100): unweighted mean over the ten service products (161 expert-selected cases in all) of short structured fields (booleans, numbers within 5%, and categories or regions judged by an LLM) extracted from the product, scored per field; direct prompting with the product artifacts, one response per task; GPT-5.2 is also a subject; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 36.13 | #105 |
| 2 | Gemini 3 Pro | 36.04 | #77 |
| 3 | Claude Opus 4.5 | 27.33 | #79 |
| 4 | Llama 4 Maverick Instruct | 26.1 | #439 |
| 5 | Qwen 3 VL 32B Instruct | 25.91 | #276 |
| 6 | Grok 4 | 19.96 | #169 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=s2sservicebench-signal-comprehension · How It Works · Data refreshed daily, snapshot 2026-10-11.