Collections

Discover the best community collections!

Collections trending this week
Qwen3.8-27B on RTX 5090
SGLang b12x + all-NVFP4. +25 to +33% decode vs stock on fixed prompts. github.com/thomgardiner/sglang-sm120-nvfp4
Qwen3.8-27B on RTX 5090
SGLang b12x + all-NVFP4. +25 to +33% decode vs stock on fixed prompts. github.com/thomgardiner/sglang-sm120-nvfp4
private-cloud
Self-hosted model research. Use hardware-specific collections for placement and verify runtime compatibility and measured memory.