Atlas

Benchmarks

← All benchmarks

Kangaroo 2025 7-8

Math · 2025-10-20

This MathArena track evaluates models on the Kangaroo 2025 7-8 problem set. Scores report the percentage of problems answered correctly.

Top models (higher is better)

ModelScore
Opus 4.896.7
Kimi K396.7
GPT-5.495.8
GPT-5.595.8
Gemini 3.5 Flash93.3
Gemini 3 Pro Preview91.7
GPT-590.8
Gemini 3.1 Pro Preview90.0
Gemini 3.6 Flash90.0
Gemini 3 Flash Preview89.2
GPT-5.289.2
Kimi K2.588.3
GPT-5 Mini87.5
Opus 4.686.7
GPT-5.185.8
Gemini 2.5 Pro82.5
Qwen3-VL-235B-A22B-Instruct82.5
Grok 480.8
Sonnet 4.580.0
Grok 4 Fast80.0
Grok 4.1 Fast79.2
GLM-4.5V75.0
Loading Atlas data…