Skip to content
Weights Not Walls
Menu

MindCube

Reported

No description yet: we only describe an eval set when we can do it correctly.

Kind
Published by a third party's leaderboard, collected by Epoch AI. The leaderboard is named and linked.
Link
https://mind-cube.github.io/
Score shown
Epoch's column "Overall score", as a percentage, higher is better. The source gives no error bars.
Results
5 in all: 1 for open-weight models, 1 for closed ones, 3 where Epoch does not say. 1 linked to a model page here.

Best open-weight result: google/gemma-3-12b-it, 46.7%, number 1 of 5 results.

MindCube results
#ModelWeightsSettingScoreDateSource
1google/gemma-3-12b-itEpoch: gemma-3-12b-it, Google DeepMindopendefault46.7%not datedmind-cube.github.io
2mPLUG-Owl3-7B-241101not stateddefault44.9%not datedmind-cube.github.io
3claude-sonnet-4-20250514_unknownAnthropicclosedeffort unknown44.8%not datedmind-cube.github.io
4LLaVA-Video-7B-Qwen2not stateddefault42.0%not datedmind-cube.github.io
5LongVA-7Bnot stateddefault29.5%not datedmind-cube.github.io

Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method