Skip to main content
Isaac 0.2 is the second generation of Perceptron’s open-weights image-based vision-language family. Ask questions about images, detect objects, read text, get captions — through the hosted API or by self-hosting the weights from Hugging Face. The 0.2 family ships in two sizes:

Try Isaac 0.2 in 30 seconds

Create an API key

Get your key from the Perceptron platform

Join Discord

Get help and see what others are building

Get started with Image

Step through this example interactively
Or pick your preferred method:
Using Python? Install with pip install perceptron or pip install openai. To trigger reasoning or grounding via raw chat completions, use <hint>THINK</hint>, <hint>BOX</hint>, <hint>POINT</hint>, or <hint>FOCUS</hint> as a system-role message.
Supported image formats: JPEG, PNG, WebP — pass a URL or local file path. Outputs are deterministic by default (temperature defaults to 0.0). See API Reference for all parameters.

Explore Isaac 0.2 capabilities

Image Q&A

Ask questions about images and get grounded answers

Object Detection

Locate targets with precise bounding boxes

OCR

Extract text from images and documents

Image Captioning

Generate descriptions of images

In-Context Learning

Adapt Isaac 0.2 to image tasks with a handful of examples

Structured Outputs

Constrain replies to Pydantic, JSON schemas, or regex

Models overview

isaac-0.2-2b-preview

Best-in-class open-weights 2B VLM with reasoning. Sub-200ms time-to-first-token.

isaac-0.2-1b

Compact 1B VLM with reasoning, optimized for edge and low-latency deployments.