LS Jilang
A tool for hybrid inference, developed by the Bilibili creator "Aitiaoshengde Nailong" (Jump-Roping Milk Dragon).
Timeline (last 90 days)
The uploader initially studied and modified LS济浪 and LKMo developed by Bilibili uploader "爱跳绳的奶龙", and validated them on an RTX Pro 5000; encountering a bottleneck because LKMo is closed-source, the uploader later rebuilt Xtier based on the official SGLang distribution.
On the uploader's RTX Pro 5000 platform, the speed performance of each model on Xtier is: Qwen 3.8 is the fastest with essentially no waste; DeepSeek 5.3 is slightly slower by 2% than the revised LS济浪; GLM 5.3 is faster than LS济浪; MiniMax M3.1 Flash is very slow.
All Facts (2)
On the uploader's RTX Pro 5000 platform, the speed performance of each model on Xtier is: Qwen 3.8 is the fastest with essentially no waste; DeepSeek 5.3 is slightly slower by 2% than the revised LS济浪; GLM 5.3 is faster than LS济浪; MiniMax M3.1 Flash is very slow.
50%UnverifiedThe uploader initially studied and modified LS济浪 and LKMo developed by Bilibili uploader "爱跳绳的奶龙", and validated them on an RTX Pro 5000; encountering a bottleneck because LKMo is closed-source, the uploader later rebuilt Xtier based on the official SGLang distribution.
50%