Individual experiment result
GermEval NER (DE)
NVIDIA Nemotron 3 Ultra 550B-A55B NVIDIA
Micro-F1 excluding MISC reasoning off
80.3%
Micro-F1 excluding MISC
Run details
- Date
- 2026-08-12
- Test cases
- 1,024
- Median latency
- 1.38s
- Total cost
- —
- Avg. prompt tokens
- 1,183.3
- Avg. answer tokens
- 25.7
- Avg. reasoning tokens
- —
- Quantization
- Unverified
About this benchmark
Native German named-entity recognition (person/location/org/misc) emitted as JSON; seqeval micro-F1 excluding MISC.
Provenance
- Canonical model
- nvidia/nemotron-3-ultra-550b-a55b
- Run ID
2026-08-12T09-31-55__openrouter__nvidia__nemotron-3-ultra-550b-a55b:free__germeval_de