gemma-4-31B-it
Performancebenchmark · gemessen am 27.07.2026 22:25
Benchmark-ID
run-20260728-034052-87f00bTimebench 3 - Kombi (Prefill + Generation)Dense31BRuntime: llama.cppQuantisierung: Q4_K_M
Generation247,60tok/s
Prefill1.828,72tok/s
Time to First Token9.212,00ms
Gesamtdauer87,30s
Concurrency5parallel
