Local natural-language-to-shell command generator. A 941 MB fine-tuned Qwen2.5-Coder-1.5B running on CPU in ~1s.

Hand-written NVFP4 W4A16 CUDA kernels and chain-MTP speculative serving — Qwen3.6-27B at up to 366 tok/s on four Tesla V100s, hardware with no FP4 support

A tool to unlobotomize your NVIDIA card!

Caching experts made model agnostic for llama. Drastically speeds up generation speed on MoE models that fully fit in the RAM but partially in Vram. WIP: Same speedup but for models that don't fit …

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

Run MoE models bigger than your RAM. A 284B on a 12 GB phone, CPU only, lossless, on stock llama.cpp

cyberneurova-DeepSeek-V4-Flash-abliterated-aligned local inference engine with M5 Metal support

Vane is an AI-powered answering engine.

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & En…

Run full Kimi K3 on a single device. And an OpenAI-compatible API server for local chat and coding agents.

Python-based stock analysis tool that combines traditional technical analysis with AI prediction capabilities. Providing comprehensive stock analysis and forecasting using K-line charts, technical …

AirLLM 70B inference with single 4GB GPU

DeepSeek 4 Flash and PRO local inference engine for Metal, CUDA and ROCm

MLX (Apple Silicon) port of poolside/Laguna-S-2.1 (118B-A8B MoE). Quants: pipenetwork/Laguna-S-2.1-MLX-*

Community benchmarks and scripts for running Poolside Laguna S 2.1

Giant MoE models on a single consumer GPU by streaming experts from SSD. CUDA fork of antirez/ds4: runs GLM-5.2 (743B), Tencent Hy3 (295B), and DeepSeek 4 Flash, with io_uring expert streaming, LFU…

Metal FP32 Vs BF16 Vs FP16 benchmark

Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

Infinite Worlds with Versatile Interactions

A Julia implementation of Interactive Brokers API

Differentiable Neural Computer in TensorFlow

Differentiable Neural Computer (DNC) implementation in PyTorch.

Neural Turing Machine (NTM) & Differentiable Neural Computer (DNC) with pytorch & visdom

Differentiable Neural Computers, Sparse Access Memory and Sparse Differentiable Neural Computers, for Pytorch

A TensorFlow implementation of the Differentiable Neural Computer.

Compile programs directly into transformer weights. Includes a 2D convex-hull KV cache with O(log n) inference.

Large Language Models (Transformer deep learning architecture)

Load nanoGPT-style transformers in Julia. Code ported from @karpathy's llama2.c

The most atomic way to train and run inference for a GPT in 100 lines of pure, dependency-free Julia.

Julia Implementation of Transformer models