Atlas

Benchmarks

← All benchmarks

PUMaC 2024 (CAISI 55-task set)

Math · 2025-09-30

CAISI's PUMaC 2024 set contains 55 text-only advanced high-school mathematics problems without visual diagrams. Problems cover algebra, number theory, combinatorics, and geometry, and scores report answer accuracy.

Top models (higher is better)

ModelScore
Claude Mythos Preview97.7
Opus 4.897.3
GPT-5.596.4
DeepSeek-V4-Pro96.4
GLM-5.295.5
Opus 4.695.0
GPT-5.4 Mini93.0
GPT-585.9
DeepSeek-V3.177.7
DeepSeek-R1-052872.7
Opus 469.1
gpt-oss-120b67.3
DeepSeek-R160.9
Loading Atlas data…