sudoingX / local-ai-on-your-gaming-gpu

run a 27b local ai on your gaming gpu, rtx 3060 and up, radeon rx 6600 and up in beta, 8gb or more, linux and windows, one line to install

professorpalmer / bonsai-ada-surgery

Bonsai 2 27B: 262k context (trained max) on 12 GB, 54 to 97 tok/s decode, 2x prefill, q8_0 KV recipe, grammar tool calls 9/9. Same ternary weights. Windows bundle + receipts.

sepiablue-ai / ComfyUI-H3-X2-Stream

ComfyUI nodes for MiniMax H3: INT8 X2 VAE and asynchronous NVENC saving. 5-second 1088x1920 video with audio in about 60 s on an RTX 4070 12GB, about 30% faster than the FP16 X2 VAE.