Microsoft's Mage-VL-Cowork Beats 27B Models at Desktop Control With Just 4B Parameters
Microsoft's 4B hybrid agent hits 66% on OSWorld-Verified by learning when to click a button and when to open a shell.
20 stories
Microsoft's 4B hybrid agent hits 66% on OSWorld-Verified by learning when to click a button and when to open a shell.
Cagliostro Lab's open-source anime SDXL finetune ships with 8.4M training images, a tag-based prompt grammar, and a refined Opt variant for better anatomy and color.
Tiny 14KB steering files let you drag motion, mood, and lighting up or down on MiniMax H3 video generations without any training.
A 2.6GB uncensored local model with 1M-token context, thinking mode, and tool calling that runs on laptops or even phones without internet.
Poolside's 225B Mixture-of-Experts coding model lands with open weights, 262K context, and benchmark scores rivaling frontier closed systems.
An independent 4B fine-tune of Qwen3.5 turns plain text prompts into self-contained Three.js scenes, with weights, dataset, and GGUFs public.
A classic Wide-ResNet-50-2 backbone retrained with the RACM recipe is quietly pulling 850K+ monthly downloads on Hugging Face for image classification work.
A community quantization pack turns NVIDIA's 30-billion-parameter speech model into drop-in GGUF files that run on a single consumer GPU.
A community port by the handy-computer team converts NVIDIA's multilingual Parakeet-TDT-0.6B-v3 into GGUF files for transcribe.cpp, enabling local CPU and GPU inference.
Apex Intelligence's first open model fine-tunes Qwen3.6-35B-A3B with help from an automated research system. It beats its base model on most research and agent benchmarks while activating only about 3B parameters per token.
Baidu's 3B open-source vision-language model parses hundred-page documents in a single request using a sparse attention trick that keeps memory flat.
Qdrant's ONNX port of multilingual-e5-large hits 926K downloads, giving teams a lightweight way to run a top multilingual embedder in production.
AtomicChat strips refusals from DeepSeek V4.1 Flash and ships 1.75 to 2.25-bit quants that run a 763B MoE on a single 12 GB GPU.
A 27B Qwen fine-tune distilled on 200K coding and agent trajectories, shipped as GGUF for local inference with llama.cpp, Ollama, and vLLM.
SYSTRAN's English-only base checkpoint repackages OpenAI Whisper for CTranslate2, delivering roughly 4x faster transcription at the same accuracy on modest hardware.
A community fine-tune strips safety refusals from Liquid AI's tri-hybrid LFM3-1.7B and ships it as a 1.1GB GGUF that runs fully offline.
DeepVK's new Russian sentence encoder packs 8,192-token context and Matryoshka embeddings into a 149M-parameter model that rivals systems 4x its size.
QuantStack's GGUF conversion of Wan2.2-I2V-A14B shrinks a 27B Mixture-of-Experts video model into quantizations that fit on consumer GPUs through ComfyUI.
A decade-old 11.7M parameter image classifier still pulls close to a million monthly downloads, and the reasons tell you a lot about modern vision pipelines.
InclusionAI's upgraded 560B hybrid MoE jumps to 44 on Artificial Analysis Index with sharper reasoning and agentic workflow gains.