OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Local & Self-hosted AI

Pmetal

Pmetal is a Rust machine-learning framework and application suite for Apple Silicon that supports local LLM inference, LoRA/QLoRA fine-tuning, model serving, quantization and Metal/MLX acceleration through a CLI, TUI, desktop GUI and Python API.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Pmetal — OpenViST project profile in Local & Self-hosted AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Pmetal when your work centers on running or adapting language models on Apple Silicon and you want a choice of command-line, terminal, desktop or Python workflows.

Who it is for

  • Developers building local LLM workflows on Apple Silicon.
  • Practitioners fine-tuning models with LoRA or QLoRA and testing them locally.
  • Users who want CLI, terminal or desktop controls for model management, inference and training.

Documented uses

  • Run inference with a selected model and prompt, including a LoRA adapter.
  • Fine-tune a model using a local dataset and configurable training parameters.
  • Prepare, merge, quantize and evaluate models.
  • Control a local OpenAI-compatible model server.
Before you choose

Requirements & limits

The project documents prebuilt binaries, a crates.io package and source builds. Building the optional desktop GUI requires Bun in addition to its separate build command.

  • The documented platform is macOS on Apple Silicon; hardware support lists M1 through M5 chip families.
  • The README specifies Rust 1.95 or later for the CLI/TUI source build.
  • The GUI is optional and combines in-process workflows with pages that drive the bundled CLI as a subprocess.
  • The documented server command is OpenAI-compatible; the README notes that source builds require the serve feature.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Build the CLI and terminal interface from source

Prepare Pmetal's CLI and TUI from its Rust source repository.

Before you start

  • macOS on Apple Silicon
  • Rust 1.95 or later
  • Git and Cargo

Steps

  1. Clone the Pmetal repository and change into its directory.

  2. Run the command below to build the CLI and TUI.

    Documented command
    cargo build --release

Expected outcome

A release build of the CLI and TUI; the README documents `pmetal tui` as the command to open the terminal control center.

Guide limitations

  • This setup recipe does not cover model download, training or inference.
  • The optional GUI has a separate build path requiring Bun.
Documentation 1 ↗
Source-based practical guide

Fine-tune a model and run inference with its adapter

Use the documented CLI workflow to train a LoRA adapter and provide it to inference.

Before you start

  • Pmetal CLI
  • A model identifier supported by Pmetal
  • A training dataset such as the documented train.jsonl input

Steps

  1. Run pmetal train with a model identifier, dataset path and output directory, configuring LoRA rank, batch size and learning rate.

  2. Run pmetal infer with the same base model and the resulting adapter file, then provide a prompt and the --chat option.

Expected outcome

A LoRA fine-tuning run followed by inference using its adapter.

Guide limitations

  • The README gives an example using Qwen/Qwen3-0.6B; it does not establish that every model or dataset is compatible.
  • The recipe does not specify hardware memory requirements.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
PmetalRun inference with a selected model and prompt, including a LoRA adapter.macOS, Self-hostedThe project documents prebuilt binaries, a crates.io package and source builds. Building the optional desktop GUI requires Bun in addition to its separate build command.
GeniexRun a GGUF language or vision-language model locally through llama.cpp.Self-hostedThe repository labels GenieX a developer preview and identifies a BSD-3-Clause license. Its documented target platforms are Windows ARM64, Android and Linux ARM64 on Qualcomm devices.
LlmfitIdentify models and quantizations that match available CPU, RAM, GPU and VRAM.CLI, Desktop, Linux, macOSThe repository is public and MIT-licensed. Documented installation routes include Scoop, Homebrew, MacPorts, uv, pip and pre-built binaries for Windows, macOS and Linux.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
Epistates/pmetal · Official homepage title · retained in its source language
How we review sourcesSuggest a correction