German Language LLM Index
a PeerBench project

Individual experiment result

INCLUDE (DE)

Gemma 4 12B Google

Exact match reasoning off bf16
69.1%
Exact match

Run details

Date
2026-06-11
Test cases
139
Median latency
27.64s
Total cost
Avg. prompt tokens
123.1
Avg. answer tokens
221.9
Avg. reasoning tokens
Quantization
bf16

About this benchmark

INCLUDE-DE v1.2: 107 valid German questions after a blinded semantic audit excluded 32 duplicate, wrong-keyed, non-unique, corrupted, or underspecified source rows consistently; strict final-answer parsing; primary metric exact match.

4-option multiple choice native · Native German Source ↗

Provenance

Canonical model
google/gemma-4-12b-it
Run ID
2026-06-11T14-39-12__openai__gemma4-12b-bf16__include_de_nothink