3 ms·
Can confirm. I've a single R9700 and have maxed out at 45 tokens/second on llama.cpp with Q4 Qwen 3.6 27B (with MTP)
by julianlam 24d ago
Can confirm. I've a single R9700 and have maxed out at 45 tokens/second on llama.cpp with Q4 Qwen 3.6 27B (with MTP)