What if a conversational AI could respond with an interactive chart instead of a wall of text and start delivering useful information before the full answer is even ready?

GPT-chat-latest in Microsoft Foundry now makes both possible. This update brings GPT-6-level intelligence to conversational and agentic applications, with stronger reasoning across coding, search, multilingual interactions, personalization, and task creation. It also introduces two capabilities that change what a response can look like and how it reaches the user: GenUI support for visual and interactive answers, and interleaved-thinking output that surfaces useful content earlier while work on the broader response continues.

Built for conversational and agentic applications, GPT-chat-latest brings the latest model improvements into an endpoint designed around natural, context-rich interaction. Previous updates have focused on more direct responses, improved factual reliability, more consistent behavior, and advanced reasoning for complex planning, research, and analysis. This update builds on that foundation while expanding both what the model can do and how users can interact with its responses.

Bringing GPT-6 intelligence to conversational experiences

With the release of the OpenAI's GPT-6 family in Microsoft Foundry, developers gained access to a new generation of models with greater capability and efficiency for workloads including complex problem solving, coding, knowledge work, and tool-driven workflows.

GPT-chat-latest builds on this evolution in model intelligence, bringing stronger capabilities into an experience designed specifically for conversation. The update improves how the model reasons and responds while strengthening many of the capabilities people rely on in everyday interactions:

• Stronger default intelligence: More capable responses across general reasoning and conversational tasks.
• Improved coding: Stronger behavior for experiences that need to understand, generate, or work with code.
• Better search: Improved search behavior helps the model find and work with relevant information as part of a conversation.
• Stronger multilingual behavior: Improved performance across languages supports more natural conversations for users around the world.
• More useful personalization and personal context: Improvements to personalization and personal-context retrieval help GPT-chat-latest make better use of relevant context when it is available.
• Improved task suggestions and creation: The model can better recognize opportunities to move from conversation to action, including suggesting or creating tasks when appropriate.

The result is a more capable conversational model that can do more than generate a stronger answer. It can work with context, information, and tools to help users move a task forward. And intelligence is only part of the update. GenUI and interleaved-thinking output expand what a response can look like and how it reaches the user, creating more visual, interactive, and responsive experiences.

Moving beyond text with GenUI

Sometimes a user needs to compare options or explore information visually. Other times, the most useful response is something they can interact with rather than simply read.

GenUI support enables GPT-chat-latest to deliver visual or interactive answers, giving developers more flexibility in how information is presented to users. This expands the traditional conversational interface. Instead of assuming every interaction needs to end with text, developers can build experiences where the presentation of an answer better reflects what the user is trying to accomplish.

The important shift is not simply toward more visual experiences. It is about making the interface itself part of the response. A conversational experience can move between understanding a request, working through relevant information, and presenting the result in a useful form. This builds on the multimodal and context-aware direction of GPT-chat-latest, which already supports experiences that combine text and image context.

For developers, this creates room to think beyond a static chat interface. For example, when a user asks "What's the weather forecast for my trip to Paris next week?" Without GenUI, the model returns a paragraph of text listing temperatures and conditions by day. With GenUI enabled, the model can return a structured weather card component with a visual day-by-day forecast, icons, and temperature ranges that your application renders as an interactive widget. Or a request like "Show me my team's sprint progress" could produce an embedded chart the user can hover over for details. The right response to a request might still be text, but it can also be a more visual or interactive experience when that format better fits what the user is trying to do.

Delivering useful information earlier with interleaved-thinking output

Complex requests can require multiple stages of work before a complete answer is ready. But the user may not need to wait for every part of that work to finish before receiving something useful. GPT-chat-latest introduces interleaved-thinking output, sometimes described as "double texting," designed to surface useful content earlier while work on the broader response continues. For instance, if a user asks "Research the top CRM tools and recommend one for a 50-person sales team," the model might first send a quick summary of the top three options it has identified, then continue working and deliver a detailed comparison with pricing and feature analysis moments later.

This creates a more progressive interaction pattern. Instead of an experience where the model appears to have nothing to say until the complete response is ready, applications can begin delivering useful information earlier in the interaction. This creates another way to think about responsiveness. The experience is not defined only by how quickly the complete answer arrives, but also by how quickly the user begins receiving something useful that can move the task forward.

Combined with GenUI, this starts to change two fundamental parts of the conversational experience: what a response can be and how that response unfolds.

From answering questions to helping users get things done

Taken individually, better coding, search, multilingual behavior, personalization, task creation, GenUI, and interleaved-thinking output address different parts of the experience. But together, they make GPT-chat-latest better suited to applications where conversation is only the starting point.

A model can work through a complex request. Search can bring relevant information into the interaction. Personal context can make the experience more relevant to the individual. Task suggestions can help move from conversation toward action. GenUI can present the result in a format suited to the task. And interleaved-thinking output can make more involved interactions useful earlier as they unfold.

That combination is useful across a range of use cases already identified for GPT-chat-latest:

• Customer support and self-service
• Planning and knowledge work
• Multimodal conversational experiences
• Multi-turn retrieval-grounded assistants

Developers can access these improvements through the existing GPT-chat-latest in Microsoft Foundry, no need to redeploy a new model.

Pricing

Get started with GPT-chat-latest in Microsoft Foundry

Explore the GPT-chat-latest documentation for model details, capabilities, and availability, then try GPT-chat-latest in Microsoft Foundry to see how these updates can strengthen your conversational and agentic experiences.