Coming Soon

Parallax
Pilot

The all-new local AI agent that doesn't just chat—it executes. Powered by the Parallax 2 architecture and featuring a massive 1 million token context window, Pilot can read entire codebases, navigate file systems, and run complex tasks autonomously on your own hardware.

Explore Parallax Labs

Pilot vs. Competition

1 Million Token Context

Pilot, powered by Parallax 2 Mini, uses Gated DeltaNet architecture. This means that Pilot can run significantly more efficient than standard Transformer models. It can run its full 1 million context on approximately 5GB of VRAM.

Agentic Toolcalling

Pilot doesn't just provide an answer, it can take action, such as create or edit files, create folders, run commands, search the web, and more tasks with the Parallax App API.

100% Local, 100% Private

Pilot is designed for local on-device inference, meaning your data is safe with Pilot, and never leaves your device.

Pilot is powered by our leading Parallax 2 family of AI models. Pilot uses Parallax 2 Mini, with an extended 1 million context window for maximum understanding and capability.

See it in action

❯ pilot "I need you to create me a report on why NVIDIA GPUs are great for gaming."

[Thinking...] I need to use the web to research about NVIDIA and AMD GPUs and their advantages and disadvantages.
[Tool: web_search] Search google.com
[Tool Output] greenwaveelectronics.com/blog/amd-vs-nvidia-gpu-comparison, redswitches.com/blog/amd-vs-nvidia-gpus/, reddit.com/r/buildapc/comments/1oxqqne/which_is_better_nvidia_or_amd/
[Thinking...] I will read these articles / websites to extract information and compile the report.
[Tool: create_file] ./nvidia-vs-amd.md
[Thinking...] I will now begin to write the report, and validate it.

❯ [Agent is executing 4 tool calls...]

ParallaxPilot

Parallax
Pilot