Scalably/moe-vs-dense-vllm-rtx5090
Viewer • Updated • 141 • 30
Scalably open data
Measured serving runs, fine-tuning results and research evidence from a one-founder AI company.
Measured serving runs, fine-tuning results and research evidence from a one-founder AI company.
Measured vLLM and EXL3 serving runs on 2x RTX 5090