Individual experiment result
GermEval NER (DE)
Gemini 3.1 Pro Google
Micro-F1 excluding MISC reasoning on / locked provider-internal
87.3%
Micro-F1 excluding MISC
Run details
- Date
- 2026-06-09
- Test cases
- 1,024
- Median latency
- 4.86s
- Total cost
- $10.04
- Avg. prompt tokens
- 1,059.7
- Avg. answer tokens
- 29.5
- Avg. reasoning tokens
- 406.7
- Quantization
- provider-internal
About this benchmark
Native German named-entity recognition (person/location/org/misc) emitted as JSON; seqeval micro-F1 excluding MISC.
Provenance
- Canonical model
- google/gemini-3.1-pro-preview
- Run ID
2026-06-09T18-27-42__vertex_ai__gemini-3.1-pro-preview__germeval_de