open-weight

A TTS model capable of generating ultra-realistic dialogue in one pass.

TTS model capable of streaming conversational audio in realtime.

The complete post-training stack for LLMs, plus RL for any problem. SFT, DPO and multi-turn agentic RL, with evolutionary hyperparameter optimization built in.

[ICLR 2025] Official Repository for "Tamper-Resistant Safeguards for Open-Weight LLMs"

Open-source, self-hosted AI workspace for local and open-weight LLMs with vLLM, LiteLLM, autonomous agents, MCP tools, deep research, artifacts, and BYOK.

What does gpt-oss tell us about OpenAI's training data?

Find what open-weight Large Language Model (LLM) can fit into your hardware and run it

NVFP4 LLM inference tuned for consumer Blackwell GPUs. Auto-detects your GPU, downloads the model, serves an OpenAI-compatible API — benchmarked and quality-certified.

FastMCP 3.2 server plus Webapp for using the leaked Claude Code locally with ollama and open weight models. Kairos and cat included

Use Cline's OpenAI-compatible models in Copilot Chat. ClinePass: DeepSeek V4, MiMo V2.5, Qwen3.7, Kimi K2.7, GLM 5.2, MiniMax M3. One API key, $9.99/mo, 2-5x rate limits. Bring your own key to native GitHub Copilot Chat.

No Hopper architecture (RTX 5090, etc.) required! <16 GB VRAM, Windows.

Utilities for quantizing and evaluating open weight models.

Demonstrator for an open-source speech-to-summary workflow

Open Weight Definition (OWD)

Open-weight coding and agent models (Qwen Coder, DeepSeek, GLM, Kimi, Devstral, StarCoder) plus the tools to run them locally: llama.cpp, Ollama, vLLM, MLX, ranked by GitHub stars

Atomic-skill compact-shell benchmark for open-weight coding agents (API Standard v1.0.1).

On-prem AI coding agent + LLM cost-router for regulated enterprises — 100% AMD, open-weight, air-gapped, Ed25519 tamper-evident audit. AMD ACT I 1st-place winner.

Run a POC on any open-weight model and validate your GPU needs in one session. Sizes VRAM, provisions the right instance, benchmarks it on vLLM, and tears everything down so you know what fits, how fast it runs, and what it costs before you commit.

BioRoute: LLM inference benchmark for healthcare agent routing — ModernBERT vs Qwen 2.5 7B (Ollama) vs Claude on F1, ECE, latency, cost, and data residency.

Sparky: open-weight AI on a USB stick. Plug into any Windows, macOS or Linux computer, choose models by what you want them for, and run them offline with nothing installed.