Run Isaac 0.1 on Replicate and Discover More

Perceptron AI has launched Isaac 0.1, a 2B-parameter, open-weight vision-language model designed for grounded perception. Isaac answers questions about images, analyzes spatial relationships, reads text in cluttered environments, and indicates where its answers originate. Despite its small size, Isaac competes with models many times larger in tasks like optical character recognition (OCR), object recognition, and visual reasoning.

One of the standout features of Isaac 0.1 is its grounded visual reasoning capability. The model not only describes a scene but also explains why its answers are correct, returning bounding boxes or regions associated with each claim. This aids in building applications requiring transparency, traceability, or step-by-step evidence.

The model showcases strong OCR performance in real-world conditions. It can read small or partially obscured text on signs, labels, packaging, and documents. Isaac combines OCR with contextual understanding, allowing users to ask questions like, “What’s the return address?” or “How much time is left in the game?”.

Isaac understands how objects relate to one another: where they are, how they interact, and when something is out of place. This makes it useful for tasks like identifying misaligned components, spotting broken parts, or determining which bin or location an item belongs to.

The model learns new tasks from examples. Show Isaac a few annotated examples of defects, components, or conditions you're interested in, and it adapts immediately without the need for fine-tuning. Built for efficiency, with just 2 billion parameters, Isaac is fast enough for real-time or edge-constrained applications.

It is practical for robotics, manufacturing, visual inspection, and document workflows at scale. To get started with the API, you can run Isaac 0.1 using JavaScript and the Replicate API as follows:

import Replicate from "replicate"; const replicate = new Replicate(); const input = { image: "https://replicate.delivery/pbxt/O3bB4rzBd1qi3wMWb1GFvjuxduAw9AfASgAkfCLcaT1380ZN/woman-street.webp" }; const output = await replicate.run("perceptron-ai-inc/isaac-0.1", { input }); console.log(output) //=> {"text":"No, it is not safe to cross the street..."}