Individual experiment result
MuSR (DE)
Nemotron 3.5 Bolt 74B-A9B (IQ2) NVIDIA
Accuracy reasoning off iq2
76.9%
Accuracy
Run details
- Date
- 2026-10-16
- Test cases
- 553
- Median latency
- 19.45s
- Total cost
- —
- Avg. prompt tokens
- 1,500
- Avg. answer tokens
- 1,917.4
- Avg. reasoning tokens
- —
- Quantization
- iq2
About this benchmark
German MuSR v1.1 defect-filtered benchmark: 553 questions (11 truncated team-allocation translations excluded), generate_until cot+; primary metric accuracy.
Subject breakdown
| Subject | Score |
|---|---|
| murder mystery | 80.4% |
| object placements | 59.4% |
| team allocation | 77.8% |
Provenance
- Canonical model
- nvidia/nemotron-3.5-bolt-74b-a9b-iq2
- Run ID
2026-10-16T03-23-21__openai__nvidia__NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB__musr_de