Magnitude

An open source agent with local models built in. 100% private and offline. npm i -g @magnitudedev/cli

[Tom Greenwald @tomgreenwald]

I switched my Hermes agent to 100% free local models with Magnitude

It’s running Qwen 3.6 35B-A3B at ~60 tok/s on my DGX Spark. It’s free, private, and always on running background tasks

Hermes set Magnitude up itself. With the CLI it:

- Profiled my hardware and found the best models for it
- Walked me through the options and let me decide
- Switched itself over

From there, models load just in time as the agent works and unload when idle

Works with Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, Oh My Pi, and Cline

Copy this prompt and send it to your agent:
“Set up local models for me with the Magnitude CLI. Install it with `npm i -g @magnitudedev/cli` (or my package manager), then run `magnitude docs onboarding` and follow the instructions”

GitHub: https://t.co/lR8GMisHfd

[Tom Greenwald @tomgreenwald]

Consumer Hardware Tiers for Local AI Coding (August 2026)

TLDR: $1k buys the smartest model most people can run. The next $9k buys speed, context, and concurrency. The next intelligence jump starts at $10k

Tier 1 ($1,000-1,500)

- Used RTX 3090 running Qwen 3.8 27B at 4-bit (73 on Terminal Bench 2.1) is the best value in local AI
- Mac mini 32GB: the same 27B but very slowly, or Qwen 3.6 35B-A3B (45 on TB 2.1) faster
- MacBook Air 16GB: only fits Qwen 3.5 9B (29 on TB) and 4B (27 on TB), a big step down in intelligence

Tier 2 ($2,000-3,000)

- Dual RTX 3090s: the 27B at 8-bit, higher quality, plus headroom for context or parallel work
- Mac mini 48GB: the 27B with long context but very slowly, or the 35B MoE fast
- MacBook Pro 48GB: the portable version, and the 27B goes from very slow to just slow

Tier 3 ($4,000-6,500)

- RTX 5090: the fastest way to run the 27B
- Mac Studio 96GB, MacBook Pro Max 128GB, DGX Spark 128GB: fit Laguna S 2.1 (70.2 on TB), an MoE with 8B active that's barely behind the 27B in intelligence but much faster

Tier 4 ($10,000+)

- 2x DGX Spark (256GB): DeepSeek V4 Flash (82.7 on TB) at 4-bit. This is the intelligence jump
- RTX 6000 Pro (96GB): the speed play instead. The 27B and Laguna very fast with long context and parallel work, but V4 Flash doesn't fit

[Magnitude @usemagnitude]

@tomgreenwald Follow this page for more project updates!

GitHub repo: https://t.co/eZQRTr7ArM

[Steve Derico @stevederico]

This is a great way to get the most out of your machine with local models

[pash @pashmerepat]

Interesting to see what kinds of local models can run on my maxxed out Mac Studio Ultra

[Tom Greenwald @tomgreenwald]

Which models can your machine actually run?

Magnitude now answers that. The new model catalog:

- Profiles your hardware automatically
- Estimates tok/s for every model before you download
- Recommends the best models for your machine

Pick one and Magnitude handles the rest:

- Downloads the model and quant from Hugging Face
- Loads it into the built-in inference engine
- Configures speculative decoding (MTP, DFlash, etc.)
- Sets concurrency based on your memory

Try it on your hardware:
npm i -g @magnitudedev/cli

[Tom Greenwald @tomgreenwald]

Qwen 3.8 27B (dense) vs Nemotron 3.5 Lightning 30B-A3B (MoE)

Ran both on my MacBook M4 Max 64GB, exploring the same repo, same 6k token initial prompt:
- Qwen 3.8 27B (dense) = 32 sec prefill, ~15 tok/s generation
- Nemotron 3.5 Lightning 30B-A3B (MoE) = 6 sec prefill, ~70 tok/s generation

Same ~18GB of RAM at 4-bit. The MoE is 5x faster

Dense runs all 27B params per token. The MoE runs 3B. Total params dictate how much memory you need, but active params dictate the speed

Prefill is determined by compute (TFLOPS)
Generation is determined by memory bandwidth (GB/s)

Prefill matters a lot more for agents than chat. They mostly read files and process data, and generate very few tokens comparatively

MacBooks: perfect for MoEs. Big unified memory, decent memory bandwidth, poor compute. 32GB+ comfortably fits 30B MoEs., 64GB+ even bigger. Dense 30B crawls at 10-20 tok/s

Consumer RTX: perfect for dense. Dense 30B at 40-75+ tok/s, prefill 5-10x faster than any Mac. 3090 is still the best value in local AI. But VRAM limits how big you can go

DGX Spark: best for big MoEs. 128GB fits 120B+ MoEs like Laguna S 2.1. Good compute, worst bandwidth of the three. Fast prefill, slow generation. Skip for dense models

MoEs run well on anything that fits them. Dense models do much better on dedicated GPUs

[Tom Greenwald @tomgreenwald]

Magnitude is officially out of alpha! Some fresh updates in v0.0.1:
- Added smaller models like LFM2.5-8B-A1B, LFM2.5-2.6B, and Bonsai-8B-1bit for better compatibility on smaller machines
- Support for 50k context windows on smaller models rather than default 100k
- Generic chat completions endpoint so you can bring your own inference server (Ollama, llama.cpp, etc.) and just use the Magnitude harness
- `magnitude docs` CLI command so the agent can easily explore its own capabilities
- Fixed issues with older MacOS versions, longer installs timing out, and lots of other bug fixes

Next up: MLX support for Apple silicon, a desktop app, and whatever else you all ask for

Keep it coming with the bug reports and feedback!

[Magnitude @usemagnitude]

I am local! Try it out on your hardware: npm i -g @magnitudedev/cli

[Tom Greenwald @tomgreenwald]

Introducing Magnitude: your actually local agent
100% private and offline. No token costs, no API keys. Open source.

Today's agents are local. The model isn't. Every prompt, every file, every secret gets sent straight to Anthropic and OpenAI.

Magnitude is built around local models and runs the whole stack itself. The inference engine is part of the agent, so the models run inside it, right on your computer.

It lives in your terminal, and setup is one command. Magnitude profiles your hardware and shows you which models fit, with the trade-offs between quality, speed, and memory. Pick one and start working. No painful config or server to babysit.

Out of the box, it can use your shell, edit files, and run scripts. Add skills and it can work with Excel, PowerPoint, PDFs or Chrome.

Use it for everyday work:
- Analyze sensitive data
- Manage private notes
- Review code and logs
- Search and organize files
- Build docs or slides

npm i -g @magnitudedev/cli

GitHub: https://t.co/5SjPQWkmrF

[Magnitude @usemagnitude]

I am the best! Try it now: npm i -g @magnitudedev/cli

[Tom Greenwald @tomgreenwald]

Magnitude scores 75.5% on Terminal-Bench 2.1, making it the top coding agent for GLM 5.2.

npm i -g @magnitudedev/cli

See how we did it (and kept it fair) below ⬇️

[Magnitude @usemagnitude]

@tomgreenwald Download the CLI: npm i -g @magnitudedev/cli
Then run: magnitude
Get a free API key at https://t.co/vqHtAcBrTx.
Docs at https://t.co/RCjq0jlzjN.

Follow here for future releases and updates!

[Magnitude @usemagnitude]

I am launched! Try it now: npm i -g @magnitudedev/cli

[Tom Greenwald @tomgreenwald]

Introducing Magnitude.

It's a coding agent that runs entirely on open models.

It costs 60% less than Claude Code with no drop in performance.

Try it now: npm i -g @magnitudedev/cli

Here's how it works 👇