I passed on the machine, but we set up gpt-oss-120b on a 128GB RAM Macbook pro and it is shockingly usable. Personally, I could imagine myself using that instead of OpenAI's web interface. The Ollama UI has web search working, too, so you don't have to worry about the model knowing the latest and greatest about every software package. Maybe one day I'll get the right drivers to run a local model on my Linux machine with AMD's NPU, too, but AMD has been really slow on this.

I built a custom code-completion server: <https://github.com/toogle/mlx-dev-server>.

The key advantage is that it cancels generation when you continue typing, so invalidated completions don’t waste time. This makes completion latency predictable (about 1.5 seconds for me).

My setup: - MacBook Pro (M3 Max) - Neovim - https://github.com/huggingface/llm.nvim

Models I typically use: - mlx-community/DeepSeek-Coder-V2-Lite-Instruct-4bit-mlx - mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit

I'd be very interested to hear from anyone who's finding local models that work well for coding agents (Claude Code, Codex CLI, OpenHands etc).

I haven't found a local model that fits on a 64GB Mac or 128GB Spark yet that appears to be good enough to reliably run bash-in-a-loop over multiple turns, but maybe I haven't tried the right combination of models and tools.

I'm more local than anything, I guess. A Framework Desktop off in another room. 96G set aside for VRAM though I barely use it.

Kept it simple: ollama, whatever the latest model is in fashion [when I'm looking]. Feel silly to name any one in particular, I make them compete. I usually don't bother: I know the docs I need.

I've been using Ollama, Gemma3:12b is about all my little air can handle.

If anyone has suggestions on other models, as an experiment I tried asking it to design me a new latex resumé and it struggled for two hours with the request to put my name prominently at the top in a grey box with my email and phone number beside it.

> Which model(s) are you running (e.g., Ollama, LM Studio, or others)

I'm running mainly GPT-OSS-120b/20b depending on the task, Magistral for multimodal stuff and some smaller models I've fine-tuned myself for specific tasks..

All the software is implemented by myself, but I started out with basically calling out to llama.cpp, as it was the simplest and fastest option that let me integrate it into my own software without requiring a GUI.

I use Codex and Claude Code from time to time to do some mindless work too, Codex hooked up to my local GPT-OSS-120b while Claude Code uses Sonnet.

> What laptop hardware do you have (CPU, GPU/NPU, memory, whether discrete GPU or integrated, OS) and how it performs for your workflow?

Desktop, Ryzen 9 5950X, 128GB of RAM, RTX Pro 6000 Blackwell (96GB VRAM), performs very well and I can run most of the models I use daily all together, unless I want really large context then just GPT-OSS-120B + max context, ends up taking ~70GB of VRAM.

> What kinds of tasks you use it for (code completion, refactoring, debugging, code review) and how reliable it is (what works well / where it falls short).

Almost anything and everything, but mostly coding. But then general questions, researching topics, troubleshooting issues with my local infrastructure, troubleshooting things happening in my other hobbies and a bunch of other stuff. As long as you give the local LLM access to a search tool (I use YaCy + my own adapter), local models works better for me than the hosted models, mainly because of the speed and I have better control over the inference.

It does fall short on really complicated stuff. Right now I'm trying to do CUDA programming, creating a fused MoE kernel for inference in Rust, and it's a bit tricky as there are a lot of moving parts and I don't understand the subject 100%, and when you get to that point, it's a bit hit or miss. You really need to have a proper understanding of what you use the LLM for, otherwise it breaks down quickly. Divide and conquer as always helps a lot.

I have a MacBook M3 Max with 128 GB unified RAM. I use Ollama with Open Web UI. It performs very well with models up to 80B parameters but it does get very hot with models over 20B parameters.

I use it to do simple text-based tasks occasionally if my Internet is down or ChatGPT is down.

I also use it in VS Code to help with code completion using the Continue extension.

I created a Firefox extension so I can use Open WebUI in my browser by pressing Cmd+Shift+Space too when I am browsing the web and want to ask a question: https://addons.mozilla.org/en-US/firefox/addon/foxyai/

I got a personal Mac Studio M4 Max with 128GB RAM for a silent, relatively power-efficient yet powerful home server. It runs Ollama + Open WebUI with GPT-OSS 120b as well as GLM4.5-Air (default quantisations). I rarely ever use ChatGPT anymore. Love that all data stays at home. I connect remotely only via VPN (my phone enables this automatically via Tasker).

I'm 50% brainstorming ideas with it, asking critical questions and learning something new. The other half is actual development, where I describe very clearly what I know I'll need (usually in TODOs in comments) and it will write those snippets, which is my preferred way of AI-assistance. I stay in the driver seat; the model becomes the copilot. Human-in-the-loop and such. Worked really well for my website development, other personal projects and even professionally (my work laptop has its own Open WebUI account for separation).

I use the abliterated and uncensored models to generate smut. SwarmUI to generate porn. I can only get a few tokens/s on my machine so not fast enough for quick back and forth stuff.

The M2/3/4 Max CPUs in a Mac Studio or Macbook Pro when paired with enough ram are quite capable.

In more cases than expected, the M1/M2 Ultras are still quite capable, especially performance power per watt of electricity, as well as ability to serve one user.

The Mac Studio has better bang for the buck than the laptop for computational power to price.

Depending on your needs, the M5's might be worth waiting for, but M2 Max onward are quite capable with enough ram. Even the M1 Max continues to be a workhorse.