Individual experiment result
MuSR (DE)
Claude Opus 4.8 Anthropic
Accuracy reasoning off provider-internal
86.3%
Accuracy
Run details
- Date
- 2026-06-13
- Test cases
- 564
- Median latency
- 13.87s
- Total cost
- $23.74
- Avg. prompt tokens
- 3,039.8
- Avg. answer tokens
- 1,075.6
- Avg. reasoning tokens
- —
- Quantization
- provider-internal
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 | 87.6% |
| object placements | 79.7% |
| team allocation | 86.8% |
Provenance
- Canonical model
- anthropic/claude-opus-4.8
- Run ID
2026-06-13T18-12-02__openai__cc__claude-opus-4-8__musr_de