Parallax
Labs

Creating amazing AI models that are highly capable, yet tiny, with a focus on reasoning capabilities as well as coding.

Try Parallax 2 Mini

Introducing Parallax 2

Our flagship reasoning model, with 9 billion parameters. Trained on over 150,000 quality reasoning, coding, and debugging traces. Experience top-tier Claude Opus reasoning locally, tool calling via MCP, and more. It is comparable to models like Claude Sonnet and GPT-5.6. What's the best part? Parallax 2 can run all locally.

Coming in ? hours...

Introducing Parallax App

Our simple, yet powerful local-first AI workbench. Chat with Parallax 2, build complex projects with Parallax 2 Agent, and chat with Parallax 2 in Chat Mode.

Available Soon

The Parallax 2 Family

Parallax 2 Mini Preview

0.8B

The smallest model in the family, running on laptops with just 2GB of RAM.

Available Now

Parallax 2 Preview

A scaled-down version of the flagship model, featuring structured Opus-style reasoning.

Available Now

Parallax 2

The flagship reasoning model. 9 billion parameters, trained on 150K+ frontier traces.

Coming Soon

PARALLAX NEWSROOM

October 19, 2026

Parallax 2 Training Run Complete

We have officially finished training the flagship Parallax 2 model on ~160,000 frontier reasoning and coding traces.

We are currently preparing to launch the weights and GGUF onto Hugging Face by tomorrow.

October 19, 2026

Introducing Parallax App: Our local-first AI mini-lab, just for you!

We are building the simplest, easiest way to enjoy local AI, Parallax App. Parallax App is a unified desktop application that combines a lightning-fast Chat interface with an autonomous Agentic coding.

Parallax App empowers you to build multi-file complex project structures using Parallax Agent, or chat and reason with Parallax Chat.

October 17, 2026

The $32 Frontier: How We Trained an Incredible Model on a Budget

While OpenAI and Anthropic are busy spending billions on GPUs, datacenters, and more... we rented a single RTX PRO 6000 for $32. By curating 160,000 elite traces from Opus, Fable, and OpenSWE, we distilled frontier-level intelligence into a 9B model that runs on consumer hardware.