Theia AI and Theia IDE Now Support GPT-5—Out of the Box!

Developers and tool builders can now use OpenAI’s GPT-5 directly in Theia AI and the AI-powered Theia IDE, without additional integration work. Just add "gpt-5" (or its variants like mini or nano) to your model list—and you’re ready to go.

If you’re new to Theia AI or the Theia IDE with built-in AI, check out the Theia AI introduction and AI Theia IDE overview, or download the AI-powered Theia IDE here.

For background on why Theia AI can easily integrate models like GPT-5, see Why Theia supports any LLM.

What We Expect from GPT-5

OpenAI released GPT-5 on August 7, 2025, describing it as its most capable model to date. While public benchmarks and hands-on experiences are still emerging, here are the main capabilities we expect and are looking forward to evaluating in practice:

• Adaptive model selection – GPT-5 internally routes requests to different specializations (fast, reasoning, mini, nano) to balance speed and depth automatically.
• Improved coding performance – We anticipate fewer hallucinations, better handling of large codebases, and improved debugging compared to GPT-4.
• Better tool use orchestration – Expected to more reliably decide when and how to call tools, sequence multiple tools in one reasoning chain, pass richer arguments, and combine results effectively. This could lead to more accurate context gathering, fewer wasted calls, and smoother multi-step workflows in Theia agents.
• Stronger grounding – Expected to provide more cautious, transparent answers when uncertain, reducing overconfident wrong responses.
• Multimodal processing – Support for text, image, audio, and possibly video inputs, with much larger context windows (reports suggest 256k tokens or more).
• Customizable interaction – Options to adjust tone or style, which we will explore for developer-facing agents.
• New developer controls – Parameters such as verbosity or reasoning_effort could allow more fine-tuned behavior in coding and tool-integration workflows.

We’re particularly interested in testing how GPT-5 performs in Theia IDE agent workflows, including code completion, refactoring, and multi-file reasoning, and in evaluating its multimodal abilities for non-code-centric tooling scenarios.

Configuring and Using GPT-5 in Theia IDE

Thanks to Theia AI’s open architecture, adding GPT-5 is as simple as updating your settings—no code changes required. Once configured, GPT-5 is instantly available to all AI agents in the Theia IDE, such as the Theia Coder.

In the demo below, you’ll see how to:

1. Configure GPT-5 in the Theia IDE so it’s ready for use.
2. Immediately put GPT-5 to work, adding itself to the default model list for future releases.

With just a few steps, you can go from zero to coding with GPT-5—fully integrated into your workflows.

Wrapping Up

With GPT-5 integrated out of the box, Theia AI and the AI-powered Theia IDE continue to make it easy to evaluate and adopt the latest AI models. As we gain more practical experience with GPT-5, we’ll share insights on its strengths, limitations, and best uses in tool-building and IDE workflows.

For more background, see our previous posts: Introducing the AI-powered Theia IDE and Introducing Theia AI.

Need help customizing your setup or building embedded AI tools? EclipseSource offers expert AI-in-tools consulting services and deep expertise in web & cloud-based tooling, Eclipse Theia and VS Code.

👉 Contact us to discuss your tool project

💼 Follow us: EclipseSource on LinkedIn

🎥 Subscribe to our YouTube channel: EclipseSource on YouTube