Simon Willison released CORS Chat, a web interface designed to test OpenAI-Responses-compatible chat endpoints. He built the tool today using GPT-5.6-Sol xhigh to evaluate Qwen 3.8 27B running in LM Studio on his M5 MacBook Pro and an NVIDIA DGX Spark. The application allows users to persist conversations in the browser and export them as JSON. It also handles SVG images by progressively rendering them while tokens stream. Willison confirmed the tool works with LM Studio using the –cors option and OpenRouter. The interface includes selectors for endpoints and models, alongside a sidebar for saved chats. A screenshot shows a conversation where the model generated an SVG of five intersecting squares on a dark navy background. The preview pane displays the overlapping shapes in pink, blue, green, yellow, and purple with white outlines. This utility addresses the technical friction developers face when testing local or remote inference services without a standardised browser experience.
- Supports SVG progressive rendering during token streaming
- Compatible with LM Studio –cors and OpenRouter backends
- Exports chat history as copy-pasted JSON files
