Condesion-Bench - Variable Conditions: leaderboard
Metric: Satisfaction rate (%) of variable conditions (which decision variables may be selected), of the single action an LLM generates zero-shot (temperature 0, JSON output) on the 251 Condesion-Bench instances (one per trading day, October 2024 to September 2025: allocate a budget over 15 S&P 500 stocks under variable, contextual and allocation conditions); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini | 94.02 |
| 2 | O3 (2025-04-16) | 92.83 |
| 3 | GPT-5 | 92.43 |
| 4 | GPT-4.1 Mini | 91.63 |
| 5 | O4 Mini (2025-04-16) | 90.84 |
| 6 | Gemini 2.0 Flash (001) | 90.84 |
| 7 | Claude 3.5 Sonnet (20241022) | 90.44 |
| 8 | GPT-OSS-120B | 90.04 |
| 9 | Claude 3.5 Haiku | 90.04 |
| 10 | GPT-4.1 | 89.64 |
| 11 | GPT-OSS-20B | 89.64 |
| 12 | Gemini 2.0 Flash Lite (001) | 88.45 |
| 13 | Llama 3.3 70B Instruct | 88.05 |
| 14 | Mistral Large 2 (Jul) | 76.89 |
| 15 | Mistral Small 3.2 | 74.5 |
Interactive version: theaggregate.ai/benchmark?slug=condesion-bench-variable-conditions · How It Works · Data refreshed daily, snapshot 2026-10-07.