Kyojin: open-source engine that runs 100 GB-class open models (GLM-5.3, Qwen3.8, DeepSeek V4.1) on one desktop machine. AMD Strix Halo, NVIDIA RTX and Apple Mac. OpenAI API, built on ExLlamaV3.
amd
apple-silicon
cuda
deepseek
exl3
exllamav3
glm
inference
llm
llm-inference
local-llm
nvidia
openai-api
quantization
qwen
rocm
strix-halo
Updated 2026-11-07 10:08:22 -10:00