Language Models

A 125-billion-parameter model that normally needs a server, packed onto one gaming PC — NVIDIA card plus RAM plus SSD — writing answers faster than you can read.

Laya exists to give multilingual documents a fast, deterministic decision layer without the latency and parsing drama of autoregressive models.

It automates the removal of transformer safety alignment so you don't have to hand-tune abliteration parameters or pay for expensive post-training.

DwarfStar bets that DeepSeek V4 Flash deserves its own self-contained C engine, complete with disk-persistent KV cache and asymmetrical 2-bit quantization, instead of yet another generic GGUF runner.

Kronos recasts noisy, multi-dimensional candlestick data as hierarchical discrete tokens so an autoregressive Transformer can forecast financial markets like a language model.

An open-source vision-language model that trains in a day and runs on modest hardware.

An open-source 30B MoE agent trained to autonomously dig through the web for answers that require more than a single search.

FinGPT offers lightweight, open-source financial LLMs and datasets as an alternative to million-dollar proprietary training runs like BloombergGPT.

Curated tutorials and tool reviews covering vibe coding, DeepSeek, Cursor, and the rest of the generative-AI menagerie, maintained as a free, open-source knowledge base.

A hosted proxy that offers free, rate-limited API access to GPT, DeepSeek, and others for Chinese users who'd rather not tunnel through a VPN.

Official Python notebooks and guides for common OpenAI API tasks.

A community-curated collection of adversarial prompts designed to bypass safety filters on major LLMs.

A rewrite of minGPT that prioritizes working, hackable training code over educational scaffolding.

It teaches how LLMs work by implementing tokenization, attention, pretraining, and finetuning in pure PyTorch, one notebook at a time.

A curated library of 60+ deep learning papers with side-by-side code and annotations, because reading arXiv shouldn't require a PhD in suffering.

It exists so you can download, run, and chat with open-weight LLMs locally through one CLI and REST API, keeping inference on your own silicon.

To give developers a single, general-purpose speech model that handles transcription, translation, and language identification by treating tasks as tokens to predict.

It is a native mobile sandbox for downloading, benchmarking, and interacting with open-source LLMs and multimodal models entirely on-device.

Because training a transformer shouldn't require 245MB of PyTorch just to multiply matrices.

WebLLM runs open-source LLMs directly inside the browser via WebGPU and exposes them through a local OpenAI-compatible API.