> ## Documentation Index
> Fetch the complete documentation index at: https://docs.perceptron.inc/llms.txt
> Use this file to discover all available pages before exploring further.

# Isaac 0.2 2B (Preview)

> 2B VLM with reasoning. Sub-200ms time-to-first-token.

## Specs

* **Model ID**: `isaac-0.2-2b-preview`
* **Inputs**: Text, Images
* **Outputs**: Text
* **Context**: 8K tokens
* **Reasoning**: Yes
* **Pricing**: \$0.15/M input, \$1.25/M output
* **Supported MIME types**: `image/png`, `image/jpeg`, `image/webp`
* **Source**: [Open weights on Hugging Face](https://huggingface.co/PerceptronAI/Isaac-0.2-2B-Preview)

## Trigger thinking & grounding

Isaac 0.2 uses `<hint>...</hint>` values inside a system-role message. Combine multiple hints in one tag, space-separated.

```json theme={null}
{
  "model": "isaac-0.2-2b-preview",
  "messages": [
    { "role": "system", "content": "<hint>BOX THINK</hint>" },
    { "role": "user", "content": [...] }
  ]
}
```

Supported hints: `BOX`, `POINT`, `POLYGON`, `THINK`, `FOCUS`. See the [API reference](/isaac-0.2/api-reference/endpoint/chat-completions) for full examples.

## Benchmarks

<img src="https://mintcdn.com/perceptron/HENOW0pHtYR0BLyy/images/isaac_0_2_benchmark_comparison.png?fit=max&auto=format&n=HENOW0pHtYR0BLyy&q=85&s=648c9344e9b76a76e5b27617816b30a4" alt="Isaac 0.x benchmark comparison" width="1500" height="1491" data-path="images/isaac_0_2_benchmark_comparison.png" />