6 ms·
qwen 3.6 27b and qen35b a3b work like magic, if we get dpark speculative decoding versions of these models it will further improve the throughput
by agenticup 3mo ago
qwen 3.6 27b and qen35b a3b work like magic, if we get dpark speculative decoding versions of these models it will further improve the throughput