Uploaded March 2026 | Updated September 2026, 2 weeks ago

---

No API keys. No cloud. No per-token cost. Just your Mac.

In this video, I show you how to run Arcee AI's Trinity Mini (26B parameters, 3B active) locally on Apple Silicon using MLX, and wire it up to OpenCode as a fully local AI coding assistant. Everything runs on-device — ideal for air-gapped environments, regulated industries, or anyone who wants a private coding AI.

I cover the full setup: choosing the right quantization for your hardware, benchmarking generation speed across all quantizations, and configuring OpenCode to talk to the local model. I also check Trinity's results with Claude Code!

⭐️⭐️⭐️ More content on Substack at airealist.ai ⭐️⭐️⭐️

*** MODELS
Arcee Trinity Mini (base) → huggingface.co/arcee-ai/Trinity-Mini
Trinity Mini MLX 8-bit → huggingface.co/mlx-community/Trinity-Mini-8bit
Trinity Mini MLX 6-bit → huggingface.co/mlx-community/Trinity-Mini-6bit
Trinity Mini MLX 4-bit → huggingface.co/mlx-community/Trinity-Mini-4bit
Trinity Mini on OpenRouter (free) → openrouter.ai/arcee-ai/trinity-mini:free

*** CODE
Full walkthrough + scripts → github.com/juliensimon/arcee-demos/tree/main/trinity-mini-mlx

*** TOOLS
MLX by Apple → github.com/ml-explore/mlx
mlx-lm (model serving) → github.com/ml-explore/mlx-lm
OpenCode (terminal coding assistant) → opencode.ai
Arcee AI → arcee.ai