Atlas

Benchmarks

← All benchmarks

MGSM German

Math · 2022-10-06

German language split of the MGSM multilingual grade-school math benchmark.

Top models (higher is better)

ModelScore
Opus 4.596.0
Gemini 3 Flash Preview95.2
GPT-5.295.2
GPT-594.4
GPT-5 Mini94.4
O4 Mini94.4
Opus 4.194.0
Sonnet 4.594.0
Gemini 3 Pro Preview94.0
Haiku 4.593.6
Opus 493.6
GPT-5.193.6
MiniMax M2.193.6
Mistral Medium 393.6
O193.6
o3-mini93.6
Qwen3-235B-A22B93.6
Claude 3.5 Sonnet (Oct 2024)93.2
Claude 3.7 Sonnet93.2
DeepSeek-V393.2
Gemini 2.5 Flash Preview (09-2025)93.2
GLM-4.793.2
GPT-4.1 Mini93.2
gpt-oss-20b93.2
Llama 4 Maverick Instruct93.2
o393.2
Qwen3 Max (rolling alias)93.2
Command A92.8
Sonnet 492.8
DeepSeek-V3.292.8
GLM-4.592.8
gpt-oss-120b92.8
Grok 3 Mini92.8
GPT-4.192.4
Llama 3.3 70B Instruct92.4
Magistral Small 1.292.4
Qwen3 Max Preview92.4
DeepSeek-R192.0
Gemini 2.5 Flash-Lite Preview (09-2025)92.0
GLM 4.692.0
Loading Atlas data…