German Language LLM Index
a PeerBench project

Individual experiment result

ScaLA (DE)

Gemma 4 31B Google

Matthews correlation coefficient reasoning off bf16
71.0%
Matthews correlation coefficient

Run details

Date
2026-06-11
Test cases
2,048
Median latency
1.28s
Total cost
$0.23
Avg. prompt tokens
910.7
Avg. answer tokens
2.5
Avg. reasoning tokens
Quantization
bf16

About this benchmark

Native German linguistic acceptability — is the sentence grammatical (ja/nein)? generate_until; primary metric MCC (accuracy recorded as secondary).

Binary acceptability native · Native German Source ↗

Provenance

Canonical model
google/gemma-4-31b-it
Run ID
2026-06-11T13-36-12__openai__gemma-4-31b-it-bf16__scala_de