Individual experiment result
GermEval NER (DE)
Gemma 4 12B Google
Micro-F1 excluding MISC reasoning off bf16
79.6%
Micro-F1 excluding MISC
Run details
- Date
- 2026-06-11
- Test cases
- 1,024
- Median latency
- 43.80s
- Total cost
- —
- Avg. prompt tokens
- 1,153
- Avg. answer tokens
- 28.1
- Avg. reasoning tokens
- —
- Quantization
- bf16
About this benchmark
Native German named-entity recognition (person/location/org/misc) emitted as JSON; seqeval micro-F1 excluding MISC.
Provenance
- Canonical model
- google/gemma-4-12b-it
- Run ID
2026-06-11T19-56-46__openai__gemma4-12b-bf16__germeval_de