Union.ai

Union.ai is an Apple Silicon accelerator for the Flyte devbox that runs machine learning tasks and apps natively on Macs using Metal.

Union.ai provides an Apple Silicon accelerator for the Flyte 2 devbox, allowing tasks and apps that request the Apple Silicon GPU to run natively on a Mac. Workloads can utilize PyTorch with MPS, MLX, and Hugging Face models directly on local hardware rather than inside a Linux container. The system bridges the local macOS process with the containerized environment, streaming logs, managing outputs, and handling app URLs. It is designed for developers iterating on machine learning code locally before deploying to a cluster. The platform also offers self-service deployment, team plans, and enterprise options with structured pricing meters.

Union.ai at a glance

Best for: Software engineer
Runs on: macOS, Linux, API, CLI, self-hosted, open source
Pricing: First $950 of usage free; Team plan available; Enterprise plan with annual commit and volume discounts; optional hosted control plane clusters at $500/month.
Released: October 2026
Category: AI coding tool
Website: union.ai

Key features

Run tasks and apps natively on Apple Silicon GPUs
Support for PyTorch MPS and MLX frameworks
Host Hugging Face models directly on local hardware
Stream host process logs and mirror exit codes
Forward app ports to the local devbox environment
Validate GPU performance with built-in proof tasks
Manage durable AI workflows and executions

Use it for

Training machine learning models locally on Mac hardware
Serving Hugging Face models from an AppEnvironment
Iterating on ML code before cluster deployment
Running concurrent training shards on a single GPU

Alternatives to Union.ai

Strata — Runs large AI models locally on consumer gaming PCs without server reliance.
oMLX — It uses paged SSD KV caching to reduce agent TTFT to under 5 seconds.
MTPLX — It runs local models twice as fast via native MTP speculative decoding.
iwant — Spins up cloud GPUs and open models instantly via a single command.
GPUStack — Unifies AI model serving and GPU compute management across heterogeneous hardware.
Foundry Local — Runs AI models locally on device with complete privacy and no Azure subscription.
TestMu AI — Autonomous AI agents and cloud execution deliver 70 percent faster test runs.
AirLLM — Runs massive 70B+ parameter models on small 4GB consumer GPUs.
fal — Access 1,000+ generative models and high-performance serverless GPUs.
AI/ML API — Access over 1000 AI models using one standard OpenAI-compatible API schema.
Cube Studio — An open source, end-to-end platform for MLOps and distributed model training.
MNN — It delivers high-performance inference and training with a lightweight footprint.

All Union.ai alternatives