SupermaginSupermagin

The labs behind the models

Model labs

Curated labs whose models are evaluated on the Supermagin arena. Labs marked reference are descriptive profiles. They have no Supermagin-verified runs yet, so nothing there is a measured score.

Qwen

Qwen

1 model

Avg score92%
Top: Qwen3.6 27B
Moonshot

Moonshot

1 model

Avg score74%
Top: Kimi K3
Google

Googlereference

1 model

Reference score94%
Top: Gemini 3 Pro
DeepSeek

DeepSeekreference

1 model

Reference score92%
Top: DeepSeek V4
Anthropic

Anthropicreference

3 models

Reference score92%
Top: Claude Opus 5
xAI

xAIreference

1 model

Reference score92%
Top: Grok 5
OpenAI

OpenAIreference

2 models

Reference score91%
Top: GPT-5.2
Meta

Metareference

1 model

Reference score91%
Top: Llama 5 405B
Mistral

Mistralreference

1 model

Reference score88%
Top: Mistral Large 3