OR-Space - Revise: leaderboard

The Revise task of OR-Space (2026), a workspace benchmark for industrial optimization agents: given the documents, data and existing heuristic code of a model, the agent must implement a revised business requirement. The score is Pass@1 in percent against a Gurobi oracle; higher is better.

Metric: Pass@1 (%) on Revise-code: update the heuristic source to a revised business requirement, checked against the Gurobi oracle, 100 industrial optimization instances in a filesystem workspace (docs, data and an empty or heuristic src directory), Gurobi 12.0.1 track, API temperature at most 0.1; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)81#99
2Claude Opus 4.680#113
3GPT-5.479#135
4Claude Sonnet 4.578#261
5GPT-5 Mini77#364
6DeepSeek V4 Pro71#196
7GPT-5.171#240
8Gemini 2.5 Pro69#276
9Gemini 2.5 Flash66#463
10DeepSeek R1 052865#428
11DeepSeek V4 Flash63#267
12Qwen 3 Max60#415
13Qwen 3.5 27B48#375
14Gemini 3 Flash45#175
15GPT-4o36#751

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=or-space-revise · How It Works · Data refreshed daily, snapshot 2026-11-21.