83.5%
Accuracy
Run details
- Date
- 2026-06-10
- Test cases
- 564
- Median latency
- 10.18s
- Total cost
- $0.26
- Avg. prompt tokens
- 1,659.8
- Avg. answer tokens
- 798.9
- Avg. reasoning tokens
- —
- Quantization
- fp8
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 |
|---|---|
| object placements | 75.0% |
| team allocation | 84.8% |
Provenance
- Canonical model
- deepseek/deepseek-v4-flash
- Run ID
2026-06-10T11-50-02__openrouter__deepseek__deepseek-v4-flash__musr_de