Condesion-Bench - Contextual Conditions: leaderboard

Condesion-Bench (2026) evaluates conditional decision-making in a compositional action space: for one trading day an LLM allocates quantities across 15 S&P 500 stocks under variable, contextual and allocation conditions, and its single zero-shot action is checked against the conditions. This board is the satisfaction rate in percent of contextual conditions, which tie feasibility to the scenario context; higher is better.

Metric: Satisfaction rate (%) of contextual conditions (feasibility tied to the scenario context), 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. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5 Mini (2025-08-07)97.21#345
2GPT-596.81#178
3O4 Mini (2025-04-16)96.41#370
4O3 (2025-04-16)96.41#231
5Claude 3.5 Sonnet (20241022)93.63#648
6GPT-OSS-120B86.06#704
7Claude 3.5 Haiku85.66#1580
8GPT-OSS-20B84.46#1183
9GPT-4.179.68#505
10Gemini 2.0 Flash (001)77.29#727
11GPT-4.1 Mini76.1#789
12Llama 3.3 70B Instruct72.51#1385
13Mistral Large 2 (Jul)71.71#1144
14Mistral Small 3.266.53#1641
15Gemini 2.0 Flash Lite (001)60.16#1118

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=condesion-bench-contextual-conditions · How It Works · Data refreshed daily, snapshot 2026-11-21.