ChatGPT begins offering interactive images and buttons with GPT-6

Summary

OpenAI is rolling out the 'Smart Interface' feature in ChatGPT alongside its GPT-6 models (Sol and Luna). This feature allows the chatbot to embed interactive diagrams, charts, forms, and clickable buttons directly within text responses.

Why it matters

This marks a shift from purely text-based conversations to an in-conversation application layer. It may reduce the need to open separate tools for calculators, visual explanations, or simple interactive tasks. The tiered model distribution (Sol for paid tiers, Luna for free/Go) also signals a strategy where OpenAI uses interface richness as a differentiator to drive subscriptions.

Highlights

• GPT-6 was trained to decide whether to generate interactive visuals or plain text. Examples include a clickable bicycle diagram and a scrollable Mahjong tile guide.
• Top-tier users access the mid-segment Sol model, while free and Go users utilize the more efficient Luna model, creating a two-tier experience.
• OpenAI claims GPT-6 shows 'stronger resistance' to attempts to bypass safety training. Additionally, improvements in web search and faster partial responses are highlighted.

Critical take

The article reads like a near-verbatim transcription of OpenAI's own blog post. There is no independent testing, no comparison with competitors' interactive features, and no discussion of failure modes or limitations. The claim of safety resistance is entirely self-reported by OpenAI with zero third-party verification. This presents as marketing rhetoric disguised as findings.