German Language LLM Index
a PeerBench project

Individual experiment result

MuSR (DE)

Tencent HY3 Tencent

Accuracy reasoning off
85.0%
Accuracy

Run details

Date
2026-08-30
Test cases
553
Median latency
8.65s
Total cost
Avg. prompt tokens
1,889.7
Avg. answer tokens
896.7
Avg. reasoning tokens
Quantization
Unverified

About this benchmark

German MuSR v1.1 defect-filtered benchmark: 553 questions (11 truncated team-allocation translations excluded), generate_until cot+; primary metric accuracy.

2–5 option multiple choice translated · German translation Source ↗

Subject breakdown

Subject Score
murder mystery 84.8%
object placements 76.6%
team allocation 87.4%

Provenance

Canonical model
tencent/hy3
Run ID
2026-08-30T19-46-08__openrouter__tencent__hy3__musr_de