Qwen3.8 27B
Qwen3.8 27B Tests Show 8-Bit KV Cache Is Worth It, 4-Bit Still Costs
Tests across 35 weight quantizations, 3 engines, and 64K-token contexts show 8-bit KV cache halves memory with little quality loss, while 4-bit drifts
1 article tagged with this topic
Tests across 35 weight quantizations, 3 engines, and 64K-token contexts show 8-bit KV cache halves memory with little quality loss, while 4-bit drifts