Atlas

Benchmarks

← All benchmarks

SAGE v1.1 - Computer Science

Math · 2026-10-27

The Computer Science subject of SAGE v1.1 (algorithms, computational problem solving, formal languages and automata), scored by class-balanced check accuracy over rubric deductions on handwritten solutions.

Top models (higher is better)

ModelScore
Gemini 4 Argon95.0
Claude Opus 5.592.8
Grok 4.792.5
GPT-6.1 Sol92.4
Claude Haiku 5.591.6
Kimi K391.6
GLM-5.3 Flash91.5
Claude Sonnet 5.591.4
Ember-191.3
DeepSeek V4.1 Flash91.1
Qwen3.7-Plus (source unspecified)90.7
Gemini 3.1 Pro Preview89.9
MiniMax M3.1 Flash89.7
Gemini 3.8 Flash89.6
Muse Spark Open89.5
GPT-5.6 Terra89.2
MiniMax M388.3
Qwen3.5 397B A17B86.7
GPT-6 Luna86.6
Gemma 4 31B IT85.9
MiMo-V2.6-Pro85.9
Qwen3.5-Flash (source unspecified)84.1
MiMo-V2.6-Flash84.0
Gemini 3.5 Flash-Lite82.7
Grok 4.1 Fast79.4
Mistral Medium 3.572.6
Mistral Small 455.1
Loading Atlas data…