OptiVerse: leaderboard
OptiVerse (2026) is a benchmark of textbook-sourced optimization problems that models solve by writing and running solver code, a problem counting as solved only when every required value matches the ground truth within 0.1%. This board is accuracy in percent over all problems; higher is better.
Metric: Accuracy (%) over all 1,000 OptiVerse problems; 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 February 2027. 19 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3 Pro | 55.9 | #145 | |
| 2 | Gemini 3 Flash | 55.5 | #175 | |
| 3 | GPT-5.2 | Thinking | 55.2 | #185 |
| 4 | Qwen3-235B-Thinking (OptiVerse checkpoint unspecified) | 51.4 | ||
| 5 | O4 Mini | 51.2 | #368 | |
| 6 | O3 | 51.1 | #237 | |
| 7 | Claude Sonnet 4.5 | Thinking | 49.7 | #261 |
| 8 | Gemini 2.5 Pro | 49.6 | #276 | |
| 9 | DeepSeek V3.2 | Thinking | 49.5 | #419 |
| 10 | Gemini 2.5 Flash | 47.4 | #463 | |
| 11 | Qwen3-30B-Thinking (OptiVerse checkpoint unspecified) | 46.3 | ||
| 12 | GPT-OSS-120B | 44.9 | #704 | |
| 13 | Qwen 3 235B Instruct (OptiVerse checkpoint unspecified) | 44.4 | ||
| 14 | Qwen 3 8B | Thinking | 39.9 | #1873 |
| 15 | Kimi K2 | 38.9 | #443 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=optiverse · How It Works · Data refreshed daily, snapshot 2026-11-21.