Public evidence index

Explore runs

Compare like with like. Every row retains its workload and measurement semantics.

25 matching runs
Try model:gemma-4 · accelerator:5090
Clear allFilters are encoded in this URL for sharing.
25 results · compact columns
SelectModel / runHardwareRuntimeWorkloadPrimary resultEvidenceStatus
nvidia/Gemma-4-26B-A4B-NVFP4NVIDIA checkpoint, concurrency 81× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.07 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Unsloth checkpoint, concurrency 81× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.6446 req/srequests/s · measured35metadata onlycompleted
RedHatAI/gemma-4-26B-A4B-it-NVFP4Red Hat AI checkpoint, concurrency 81× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.21 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Concurrency sweep, c=11× NVIDIA GeForce RTX 5090vllm 0.27.160001c=11.18 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Concurrency sweep, c=21× NVIDIA GeForce RTX 5090vllm 0.27.160001c=22.31 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Concurrency sweep, c=41× NVIDIA GeForce RTX 5090vllm 0.27.160001c=44.55 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Concurrency sweep, c=81× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.6446 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 20481× NVIDIA GeForce RTX 5090vllm 0.27.160001c=87.43 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 40961× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.38 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 61441× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.5 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 81921× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.53 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 102401× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.65 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 107521× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.58 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 112641× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.52 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 117761× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.56 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4GPU rotation validation, device 01× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.61 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4GPU rotation validation, device 11× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.58 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4GPU rotation validation, device 21× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.63 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4GPU rotation validation, device 31× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.6 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Full 500-request validation A1× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.54 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Full 500-request validation B1× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.59 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Full 500-request validation C1× NVIDIA GeForce RTX 5090vllm 0.27.160001c=88.64 req/srequests/s · measured35metadata onlycompleted
unsloth/gemma-4-26B-A4B-it-NVFP4Scheduler token budget 12288 (startup failure)4× NVIDIA GeForce RTX 5090vllm 0.27.160001c=8No result35metadata onlyfailed
unsloth/gemma-4-26B-A4B-it-NVFP4Interrupted preliminary screening4× NVIDIA GeForce RTX 5090vllm 0.27.160001c=8No result35metadata onlyinterrupted
unsloth/gemma-4-26B-A4B-it-NVFP4Derived four-replica throughput1× NVIDIA GeForce RTX 5090vllm 0.27.160001c=834.48 req/srequests/s · derived35metadata onlycompleted