Companion code for the Manning book 'LLM Customization and Fine-Tuning.' Adapt open-weights LLMs end to end: prompting and RAG, LoRA/QLoRA, full SFT, distillation, and DPO/RLHF alignment.

Structured YAML catalog of 4,587 AI models across 95 providers — pricing, context windows, modalities, capabilities. First-party data with TypeScript types and Zod validation.

Local-first desktop app for running uncensored, refusal-ablated language models. Tauri + Rust + Ollama. 12MB installer, zero telemetry.

Curated list of open-source LLM-serving runtimes, routers, evaluators, and standards. Run LLMs without locking into one vendor.

tiny tool, big videos. topic in, video out. no film degree required. webapp, windows app and mcp server for agent use

Pre-registered adversarial robustness study testing whether model-initiated session termination provides defensive coverage beyond refusal training against multi-turn attacks. Minimal Python harness with scorer and analysis pipeline. Pilot on Gemma 4 26b. Preliminary findings in FINDINGS.md.

GLM-5.2 Released: 1 Million Usable Context Window, MIT Open-Source Licensing, & Claude Code Integration Guide.

Open-weights GIS-specific model evaluation, training, and distribution assets

Collect and refine high-quality prompts for Ideogram 4.0 to improve your image generation results.

FastMCP fleet MCP server for diffusion LMs (dLLM). DiffusionGemma on Goliath RTX 4090 — batch inference, HLE-shaped reasoning, ~200–400 tok/s. Doc phase; llama-diffusion-cli sidecar next. Complements local-llm-mcp.

AI tool licenses and data terms — know what you're signing.

Production-ready multi-facet conversation evaluation benchmark - 399 facets, 5-level ordinal scoring, scales to 5000+ without redesign. Open-weights LLM (Qwen 3 8B) via Ollama/vLLM/HF.

Ideogram 4 本地 CUDA 图像生成器：内置 NF4 权重、官方 Magic Prompt 提示词优化、Windows/Linux 便携包 / Local open-weight text-to-image generator

Democratize AI: a download-and-forget app that turns idle computers into a language model owned by everyone. Folding@home, pointed at intelligence instead of proteins.

Data-driven exploration of open weight AI models and providers, based on performance x cost x energy use.

Open-weight Brazilian Portuguese LLMs trained from scratch — Apache 2.0

VoxReach POC — full-duplex AI receptionist demo built on Kyutai MoshiRAG (open weights). Persona: Hearth & Pass Korean restaurant. 2-pane investor demo UI + POS-write moat story.

Physical Intelligence — foundation models for robotics

API Evangelist profile of Mixedbread - embeddings and reranking API

Open-weights image generation models — Flux family.