Local LLM Engines and Tools: Release Tracker
Inference engines, runtimes, quantization tools and UIs for running open-weight models locally, as tracked by Local Model Watch. Each page carries the full release history compiled from GitHub, basic project facts and our articles. Version numbers and dates are collected by code, not written by an AI model.
Last updated 2026-11-03 (JST). 25 projects.
Contents
Inference engines and runtimes
| Project | What it is | Latest release | Released |
|---|---|---|---|
| llama.cpp | LLM inference engine written in C/C++ | v0.8.0 | 2026-10-28 |
| Ollama | Local LLM runtime that pulls and runs models with a single command | v0.40.13 | 2026-10-29 |
| vLLM | High-throughput LLM serving engine built around PagedAttention | v0.33.0rc3 | 2026-11-02 |
| SGLang | LLM serving framework known for RadixAttention prefix caching and structured output | v0.5.23 | 2026-11-03 |
| KoboldCpp | Single-file runtime that bundles llama.cpp with a web UI and API | v1.124 | 2026-10-31 |
| llamafile | Packages model weights and llama.cpp into one executable that runs on any OS without installation | 0.10.6 | 2026-09-16 |
| LocalAI | Self-hosted, OpenAI-API-compatible inference server that fronts multiple backends for text, image and audio | v4.12.1 | 2026-10-23 |
| MLC LLM | TVM-based compiler that deploys LLMs natively to phones, browsers (WebGPU) and a wide range of GPUs | — | — |
| RamaLama | Runs local LLMs inside OCI containers, treating models as container images on Podman or Docker | v0.25.0 | 2026-09-25 |
| exo | Framework that clusters everyday devices (Macs, PCs, phones) to run a single model across them | v1.0.71 | 2026-04-24 |
| Magnitude | Open-source inference engine written in Rust (Apache-2.0) that compiles and tunes its kernels on the device it runs on | @magnitudedev/cli@0.2.15 | 2026-10-30 |
Quantization, model formats and fine-tuning
| Project | What it is | Latest release | Released |
|---|---|---|---|
| ggml | Tensor library underlying llama.cpp and whisper.cpp | v0.28.0 | 2026-10-28 |
| ExLlamaV3 | Inference library tuned for consumer NVIDIA GPUs | v1.6.6 | 2026-10-31 |
| ik_llama.cpp | Fork of llama.cpp focused on CPU and hybrid performance, with its own quantization types (IQ*_K and others) | — | — |
| Unsloth | Library for fast, low-VRAM fine-tuning (LoRA / QLoRA) | v0.1.913-beta | 2026-10-30 |
UIs and desktop apps
| Project | What it is | Latest release | Released |
|---|---|---|---|
| Open WebUI | Self-hosted chat UI for Ollama and OpenAI-compatible APIs, with RAG, multi-user support and plugins | v0.12.5 | 2026-10-29 |
| Jan | Offline desktop chat app with llama.cpp built in; doubles as a local API server | v0.8.8 | 2026-10-28 |
| TextGen | Desktop app for local LLMs (formerly text-generation-webui) | v4.9 | 2026-05-21 |
| LM Studio CLI (lms) | Command-line tool for the LM Studio desktop app: pull, load and serve models from scripts | — | — |
Image, video and audio
| Project | What it is | Latest release | Released |
|---|---|---|---|
| ComfyUI | Node-based UI for building image, video and audio generation pipelines | v0.43.0 | 2026-10-27 |
| Diffusers | Hugging Face’s Python library for diffusion models; the standard way to run image, video and audio generators from code | v0.42.1 | 2026-10-27 |
| InvokeAI | Desktop/web app for image generation with a unified canvas and workflow editor | v6.14.3 | 2026-11-02 |
| whisper.cpp | C/C++ port of OpenAI’s Whisper speech recognition on ggml | v1.9.5 | 2026-10-07 |
| TTS WebUI | Web UI that brings many text-to-speech and music generation models under one roof | v1.5.2 | 2026-09-01 |
| Stable Diffusion WebUI (AUTOMATIC1111) | The classic web UI for Stable Diffusion, with a large extension ecosystem | v1.10.1 | 2025-02-09 |