# quantization, model, revision[, minimum capability[, safetensors load strategy]]
fp8, amd/Meta-Llama-3.1-70B-Instruct-FP8-KV, main, 80, prefetch
None, microsoft/phi-4, main, 80, prefetch
fp8, amd/Mixtral-8x22B-Instruct-v0.1-FP8-KV, main, 80, prefetch
