Individual experiment result
MuSR (DE)
GPT-5.6 Terra OpenAI
Accuracy reasoning off provider-internal
78.7%
Accuracy
Run details
- Date
- 2026-08-13
- Test cases
- 553
- Median latency
- 7.62s
- Total cost
- $2.26
- Avg. prompt tokens
- 1,445.7
- Avg. answer tokens
- 382.9
- 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 | 82.8% |
| object placements | 75.0% |
| team allocation | 75.3% |
Provenance
- Canonical model
- openai/gpt-5.6-terra
- Run ID
2026-08-13T09-55-11__openrouter__openai__gpt-5.6-terra__musr_de