OptiVerse - Combinatorial Optimization: leaderboard
Metric: Accuracy (%) on the 238 combinatorial optimization 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 | 59.24 |
| 2 | Gemini 3 Pro | 57.14 |
| 3 | GPT-5.2 (Thinking) | 57.14 |
| 4 | O4 Mini | 54.2 |
| 5 | O3 | 53.78 |
| 6 | Claude Sonnet 4.5 (Thinking) | 52.52 |
| 7 | DeepSeek V3.2 (Thinking) | 52.52 |
| 8 | Gemini 2.5 Pro | 49.16 |
| 9 | DeepSeek V3.2 (Non-reasoning) | 47.48 |
| 10 | Gemini 2.5 Flash | 46.22 |
| 11 | Kimi K2 | 44.54 |
| 12 | Qwen 3 8B (Thinking) | 43.7 |
| 13 | Qwen 3 Coder 30B A3B Instruct | 31.09 |
| 14 | Qwen 3 8B (Non-reasoning) | 22.69 |
| 15 | InternLM3-8B-Instruct | 7.56 |
Interactive version: theaggregate.ai/benchmark?slug=optiverse-combinatorial-optimization · How It Works · Data refreshed daily, snapshot 2026-10-07.