ComplexConstraints: leaderboard
Surge AI's 75 expert-written single-turn prompts from professional work such as film production and restaurant scheduling, whose constraints hinge on one another, apply only conditionally, or must be inferred from context. An LLM judge checks each prompt's 10 to 40 rubric criteria separately. The board reports the share of prompts on which every criterion is met, higher is better. Source: arxiv.org/abs/2606.09118.
Metric: Score (%). Source: surgehq.ai. Saturation forecast: Around 2029. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Max) | 57.7 | #6 (GPT-6 Astra) |
| 2 | Gemini 4 Argon (High) | 53.3 | #8 (Gemini 4 Argon) |
| 3 | Muse Spark 1.3 (xHigh) | 51.9 | #24 (Muse Spark 1.3) |
| 4 | GPT-6 Sol (Max) | 50.8 | #14 (GPT-6 Sol) |
| 5 | GPT-5.6 Sol (Max) | 50.5 | #19 (GPT-5.6 Sol) |
No result here: #7 GPT-6.1 Sol, #9 Claude Fable 5.1, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=complexconstraints · How It Works · Data refreshed daily, snapshot 2026-10-20.