The best local coding model is the largest code-tuned model your Mac's memory holds: Qwen 2.5 Coder from 8 GB up, and Qwen 3 Coder 30b from about 48 GB. Qwen 2.5 Coder comes in 3b (1.9GB), 7b (4.7GB), 14b (9.0GB) and 32b (20GB) sizes with a 32K context window. Qwen 3 Coder 30b is a 19GB download with a 256K context window, built for agentic coding, and Grux's own catalogue estimates it needs about one and a half times its download in memory, which is why it wants a 48 GB Mac. The full Qwen 3 Coder, 480b at 290GB, is beyond any Mac. All of them run through Ollama with one command. Grux OS (MIT) lists Qwen 3 Coder in its Local Models catalogue and recommends what fits the Mac it runs on; it is not a code editor, but it can hand a change to the coding agent you already use.
| Your Mac's memory | Coding model | Download |
|---|---|---|
| 8 GB | Qwen 2.5 Coder 3b | 1.9GB |
| 16 GB | Qwen 2.5 Coder 7b | 4.7GB |
| 24 GB | Qwen 2.5 Coder 14b | 9.0GB |
| 32 GB | Qwen 2.5 Coder 14b; the 32b is tight | 9.0GB, 20GB |
| 48 GB or more | Qwen 3 Coder 30b, or Qwen 2.5 Coder 32b | 19GB, 20GB |
The table applies the rule on the which LLM can my Mac run page to the sizes on Ollama's Qwen 2.5 Coder and Qwen 3 Coder pages, read on 4 October 2026.
brew install ollama
ollama run qwen2.5-coder:14b # 9.0GB, 24 GB Macs
ollama run qwen3-coder # 30b, 19GB, 48 GB Macs| Claim | Source |
|---|---|
| Qwen 2.5 Coder sizes 3b to 32b, 32K context | Ollama library, qwen2.5-coder |
| Qwen 3 Coder 30b is 19GB with a 256K context; the 480b is 290GB | Ollama library, qwen3-coder |
| Grux's catalogue lists Qwen 3 Coder 30b with its disk size and its larger memory estimate | Cookbook.swift |