Atlas

Benchmarks

← All benchmarks

BabyVision with tools

Multimodal · 2026-01-10

BabyVision with tools evaluates visual reasoning and perception with tool access. Scores are reported as percentage task performance.

Top models (higher is better)

ModelScore
DeepSeek V4.1 Flash 101792.5
Qwen3.8-Max91.3
MiniMax M3.1 Flash91.0
Claude Fable 590.5
DeepSeek V4.1 Flash89.6
GPT-5.6 Sol88.9
Kimi K385.7
GPT-5.583.6
Opus 4.881.2
Muse Spark Open80.2
Muse Spark 1.176.3
Qwen3.7-Plus (source unspecified)70.4
Kimi K2.668.5
Loading Atlas data…