🚀 Qwen-MTP Collection ⚡ MTP (Multi Token Prediction) speculative decoding enables models like Qwen3.6 to have ~1.4-2.2x faster generation with no change in accuracy. • 6 items • Updated 8 days ago • 17
MiniCPM-o & MiniCPM-V Collection Multimodal models with leading performance. • 32 items • Updated 11 days ago • 84
Running on CPU Upgrade Featured 148 Ostris' AI Toolkit 💻 148 Train FLUX, Qwen and Wan LoRAs with Ostris Ai Toolkit