Osaurus

Own your AI. Agents that remember, execute code in isolated VMs, and stay reachable from anywhere -- all on your Mac. Any model. No cloud required. Open source.

[Osaurus @OsaurusAI]

Models get commoditized.
The harness compounds.

Thanks @SarahPerezTC for the writeup.

[TechCrunch @TechCrunch]

Osaurus brings both local and cloud AI models to your Mac https://t.co/bCDBkgc3fc

[Osaurus @OsaurusAI]

Gemma 4 12B is live in Osaurus. Multimodal, runs locally through a pure Swift MLX runtime. No Python.

https://t.co/sSt5qn5hfQ

[Osaurus @OsaurusAI]

A model reading the paper that invented it.

Gemma 4 12B. From a screenshot. Local.

[Google Gemma @googlegemma]

Meet Gemma 4 12B!

A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.

Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇

[Osaurus @OsaurusAI]

@Critttr @osanseviero It's coming soon!

[Osaurus @OsaurusAI]

@qifengbuilds @atlas_cloud_ai Yes, you can tune it, and also have custom regex list you can configure.

[Osaurus @OsaurusAI]

https://t.co/sSt5qn5hfQ

[Osaurus @OsaurusAI]

@LiquidAI said device-optimized.

So I put it in airplane mode.

Read my files. Parsed a PDF. Pulled the contact. Zero internet.

LFM2.5-8B-A1B. ~100 tok/s.

[Osaurus @OsaurusAI]

All local. LFM2.5-8B-A1B in MXFP8 on Apple Silicon.

8B MoE, 1.5B active. Custom Swift MLX runtime, no Python.

Download it before your next flight.

[Osaurus @OsaurusAI]

@compileandpush @atlas_cloud_ai We're fully open, built from public since day 1. We have dedicated community members helping us manage

[Osaurus @OsaurusAI]

@Camcamera11 @atlas_cloud_ai Everything else stays local. You can point inference to a local model for full privacy. If you point to the cloud model, it will only send what is passed through inference, and you can scrub personal information using our built-in privacy filter.

[Osaurus @OsaurusAI]

@qifengbuilds @atlas_cloud_ai We have a built-in privacy filter that scrubs your private data.

[Atlas Cloud @atlas_cloud_ai]

@OsaurusAI × Atlas Cloud integration is now live.⚡️

OsaurusAI is an open-source, Swift-native AI agent framework for macOS — local-first, so your data and agent context never leave your machine.

😎And you can use Atlas Cloud's LLM API in Osaurus now.

#Osaurus #AtlasCloud #AIIntegration #Swift #macOS #API #AItool #Opensource

[Osaurus @OsaurusAI]

@MaksimXBT @MiniMax_AI compute is all they need i guess

[Osaurus @OsaurusAI]

gave @MiniMax_AI M3 one prompt:

build an idle game called "NEED MOAR COMPUTE."

it one-shotted a full frontier lab simulator. planned it, wrote it, ran it in a sandbox, tested its own code.

a model building a game about needing more compute.

[Osaurus @OsaurusAI]

Osaurus runs any model, local or cloud. https://t.co/sSt5qn5hfQ

[Osaurus @OsaurusAI]

running through Osaurus in an isolated sandbox.
real 38 KB file. playable.

[Osaurus @OsaurusAI]

Osaurus runs any model, local or cloud. https://t.co/sSt5qn5hfQ

[Osaurus @OsaurusAI]

asked it to build an idle game called "NEED MOAR COMPUTE."

one-shotted a full frontier lab simulator.

a model building a game about needing more compute.

[MiniMax (official) @MiniMax_AI]

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities

- Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
- MiniMax Sparse Attention scales context to 1M
- Natively Multimodal from Step Zero

API: https://t.co/fHRdSV7BwZ
Token Plan: https://t.co/BDCycxepZw
🚀New! MiniMax Code: https://t.co/GvB4YiB6Ul

Weights & Tech Report in ~10 Days

[Osaurus @OsaurusAI]

MiniMax M3 running through Osaurus in an isolated sandbox.
It planned the build, wrote one HTML file, ran it, and verified its own JS and HTML before surfacing it.
Real 38 KB file. Playable.

[Osaurus @OsaurusAI]

@icefrog_sol @TechCrunch Local-first everything, inference is local/cloud agnostic. There's no backend, everything is running locally.

[Osaurus @OsaurusAI]

@PsudoMike @TechCrunch You can choose both as default. Osaurus also ships with it's own router, which is hosted locally