Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

SAGE

An educational assessment benchmark evaluating LLM performance on academic and pedagogical tasks. SAGE measures models' ability to handle educational content across multiple academic levels and subject areas, testing both knowledge recall and the ability to explain concepts effectively in educational contexts.
Source:

Model Performance

#1
Claude Opus 4.5 (Thinking)
52.1%
#2
GPT-5.2
49.3%
#3
Claude Opus 4.5
45.0%
#4
Gemini 2.5 Pro (Jun 2025)
41.9%
#5
Claude Sonnet 4.5
32.9%
#6
Claude 4.1 Opus
31.4%
#7
Grok 4.20 (Reasoning)
31.3%
#8
Claude 4.1 Opus (Thinking)
30.4%
#9
Gemini 3.1 Flash Lite Preview
28.0%
#10
Grok 4.20
11.9%