Atlas

Benchmarks

← All benchmarks

PerceptionBench - Context Integration

Multimodal · 2026-07-16

The PerceptionBench context-integration category contains 255 questions that require combining visual evidence across regions or multiple images while remaining grounded in the presented content.

Top models (higher is better)

ModelScore
Gemini 3.1 Pro Preview61.2
Doubao Seed 2.1 Pro60.4
GPT-5.6 Sol60.0
Claude Fable 559.8
GPT-5.558.0
Gemini 3.5 Flash53.3
Kimi K353.3
Qwen3.7-Plus52.2
Qwen3.5 397B A17B50.2
Gemma 4 31B IT45.9
Opus 4.844.7
Kimi K2.640.4
Grok 4.539.6
GLM-4.6V33.7
GLM 5V Turbo32.9
MiniMax M331.0
Loading Atlas data…