OptiVerse - Medium: leaderboard
Metric: Accuracy (%) on the 400 medium-level problems of OptiVerse; the model writes and runs solver code (gurobi, pyomo, cvxpy, ortools and others) and a problem counts as solved only when every required variable and the objective match the ground truth within 0.1% relative error, extracted and checked by a DeepSeek-V3.2-Chat judge; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 53.25 |
| 2 | Gemini 3 Pro | 52.75 |
| 3 | GPT-5.2 (Thinking) | 50.75 |
| 4 | O3 | 47.25 |
| 5 | O4 Mini | 46.75 |
| 6 | Claude Sonnet 4.5 (Thinking) | 45.25 |
| 7 | DeepSeek V3.2 (Thinking) | 44.5 |
| 8 | Gemini 2.5 Pro | 43.75 |
| 9 | Gemini 2.5 Flash | 42.75 |
| 10 | GPT-OSS-120B | 39.25 |
| 11 | DeepSeek V3.2 (Non-reasoning) | 39.25 |
| 12 | Qwen 3 8B (Thinking) | 33 |
| 13 | Kimi K2 | 31.5 |
| 14 | Qwen 3 Coder 30B A3B Instruct | 19.25 |
| 15 | Ministral-3-8B-Instruct-2512 | 14.75 |
Interactive version: theaggregate.ai/benchmark?slug=optiverse-medium · How It Works · Data refreshed daily, snapshot 2026-10-07.