Individual experiment result
GermEval NER (DE)
DeepSeek V4 Flash 0423 DeepSeek
Micro-F1 excluding MISC reasoning off fp4
80.7%
Micro-F1 excluding MISC
Run details
- Date
- 2026-08-11
- Test cases
- 1,024
- Median latency
- 0.89s
- Total cost
- $0.11
- Avg. prompt tokens
- 1,181.2
- Avg. answer tokens
- 26.9
- Avg. reasoning tokens
- —
- Quantization
- fp4
About this benchmark
Native German named-entity recognition (person/location/org/misc) emitted as JSON; seqeval micro-F1 excluding MISC.
Provenance
- Canonical model
- deepseek/deepseek-v4-flash-0423
- Run ID
2026-08-11T12-28-07__openrouter__deepseek__deepseek-v4-flash__germeval_de