Instinct MI250X 128GB

AMD · 128GB · 2 reports

This page is thin (2 of 3 reports needed for indexing). Help fill it in.

Latest Most reported Fastest t/s

Gemma 4 25.2B (3.8B active) 26B Instruct (MoE)

Instinct MI250X 128GB · Ollama

throughput:: 35.0 t/s gen
quant:: Q4_K_M (gguf)

text-generation

Gemma 4 26B MoE on Instinct MI250X at Q4_K_M. Source: gemma4-ai.com AMD GPU guide

Gemma 4 30.7B Instruct

Instinct MI250X 128GB · vLLM

throughput:: 20.0 t/s gen
quant:: FP16 (safetensors)

text-generation

Gemma 4 31B on Instinct MI250X via vLLM + ROCm. Full FP16. Previous gen datacenter but capable. Source: gemma4-ai.com AMD GPU guide