Search

Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported

Rapid-MLX is an open-source (Apache 2.0) OpenAI- and Anthropic-compatible LLM inference server for Apple Silicon, built on MLX, focused on reliable tool calling for coding agents. Up to 4× faster than Apple's MLX (mlx-lm) on the same weights.

The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNetV4, MobileNet-V3 & V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more

Local Qwen 3.8 27B uncensored Q4_K_M + harvested SYSTEM pack. Official 3.8 weights, not a 3.6 retitle.

The AI developer platform. Use Weights & Biases to train and fine-tune models, and manage models from experimentation to production.

🔥[ACMMM 2023, Official Code] for paper "EAT: An Enhancer for Aesthetics-Oriented Transformers". Official Weights and Demos provided. 目前是地表最强开源美学评估模型之一.

RapidLLM is an LLM inference framework with continuous batching, tensor and data parallelism, quantized weights, and with pluggable Triton/CUDA kernels.

Checkpoint-engine is a simple middleware to update model weights in LLM inference engines

Noodle provides primitive modular functions for convolution layer, dense layer, pooling and activations. It allows streaming the intermediate activations, weights, and biases from/to SD/FFat/SD_MMC filesystems to overcome RAM limitations.

Sync Hevy gym workouts to Garmin Connect with exercise names, sets, reps, weights, HR overlay, and calorie estimation

Run the full 2.78-trillion-parameter Kimi K3 model, DeepSeek V4.1 Flash or GLM-5.3-Flash beyond available RAM by streaming activated weights directly from NVMe. A dependency-free, embeddable C inference engine.

Swiftlet is a Swift and Metal runtime that runs large Qwen Mixture-of-Experts models locally on Apple devices by streaming expert weights from storage, enabling 35B and 80B models to run with low RAM, including on iPhone.

Skill training with a corruption and reconstruction loop: train an agent's skill library, not its weights.

Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds. Weights on Hugging Face.

Official Hi3D v1.0: high-resolution 3D geometry refinement built on Sparc3D. Inference code, model weights, and documentation.

A computer, compiled into the weights of a language model. An exact CPU inside Qwen3.5-4B, with Gated DeltaNet state as RAM.

Complete logs of the experiments in the blog post 'LLMs can understand and write neural network weights'