Skip to main content
At Perceptron, we’re building a family of vision-language models (VLMs) that can see, reason, and act. Our VLMs allow you to:
  • Ask questions about images and videos
  • Clip videos based on specific events
  • Detect and point to objects in images and videos
  • Reason over images with tool calling
Join our Discord community to get help, offer feedback, and see what others are building.

​
Meet the models

Isaac 0.2 2B Preview

Our 2B-param image VLM for grounded perception. Reasoning enabled.

Isaac 0.2 1B

Our 1B-param image VLM for grounded perception.

Isaac 0.1 - image

Our original 2B image VLM for grounded perception.

Qwen3VL

Qwen’s large-scale VLM for visual reasoning.

Developer quickstart

Integrate Isaac into your vision stack or product.

API reference

Full API reference and examples.

Demo environment

Try prompts and API calls in the browser.

Model playground (coming soon)

Create, iterate on, and save your prompts.

​
Tutorials

Isaac 0.1 frame-by-frame

Run detections across an MP4 and rebuild an annotated video, straight from the cookbook tutorial.