SoftOR LexiBench 45 - Codegen: leaderboard
ASTOR / SoftOR (2026): 45-instance LexiBench backbone comparison, hard-constraint feasibility and lexicographic objective accuracy. Each named autoformulation method has its own board. Gemini Flash Lite is non-reasoning; GPT-5 uses medium reasoning; Gemini Pro uses provider reasoning configuration. Standard errors are not scores. Solution Accuracy (%); higher is better.
Metric: Solution Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 80 | #58 |
| 2 | GPT-5 (Medium) | 68.9 | #98 (GPT-5) |
| 3 | Gemini 3.5 Flash Lite | 44.4 | #157 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=softor-lexibench-45-codegen · How It Works · Data refreshed daily, snapshot 2026-10-21.