JOR-Bench - OptiBench: leaderboard

Metric: Accuracy (%; OptiBench textbook optimization problems including nonlinear programs, Japanese translation with human post-editing; the model writes a standalone Python program in a fenced block, run with OR-Tools available (300-second and 32 GB limits), correct when the printed value matches the reference within 1e-6; higher is better). Source: arxiv.org. Saturation forecast: Around 2029. 7 models tracked.

Top models

#ModelScore
1Qwen 3 8B62
2Qwen 3 32B58.8
3GPT-OSS-120B46.4
4Qwen3-Swallow-32B-RL-v0.241.7
5Qwen3-Swallow-8B-RL-v0.231.7
6NVIDIA-Nemotron-Nano-9B-v2-Japanese4.1

Interactive version: theaggregate.ai/benchmark?slug=jor-bench-optibench · How It Works · Data refreshed daily, snapshot 2026-09-29.