GPT-5 is Available in Factory
Starting today, GPT-5 is the default model across all droids.
Delegate with Confidence
Through our evaluations, we have found that GPT-5 is highly agentic, thorough, and brings exceptional intelligence and reasoning to every task. It is noticeably more detail-oriented than previous state-of-the-art models, particularly when searching and planning. You can hand off complex work to it and have greater trust that it will be done right.
Within Factory, GPT-5 exhibits strong built-in behaviors: it scans the context, forms a plan, executes it, tests the result, and checks that the work is complete. It follows instructions, asks sharp clarifying questions, and grounds its responses with extensive search and deep codebase understanding. Like working with an excellent software engineer, GPT-5 is opinionated—it pushes back when appropriate, suggests better approaches, and thinks through edge cases you might miss.
In Close Collaboration with OpenAI
Over the last few weeks, we've worked closely with the OpenAI team to ensure GPT-5 fits seamlessly into Factory. This collaboration builds on our long-term research and product collaboration with OpenAI, enabling models to feel native to how developers actually work in Factory.
Factory and OpenAI teams collaborated closely to make GPT-5 the default for developers.
.avif)
Where GPT-5 Works Best
Complex Refactoring
When you ask Code Droid to modernize an authentication system across your codebase, it maps out the full scope of changes, identifies clean abstraction boundaries, and systematically updates each component. You get a complete refactor with tests passing—the kind of careful work that typically consumes days.
Technical Architecture
Ask Knowledge Droid to design a system and it engages as a technical partner. It grounds its recommendations by extensively searching your codebase and integrating relevant context from documentation and memory, then weighs trade-offs against your constraints and challenges assumptions with minimal prompting. You get a design aligned to your repo and constraints, plus concrete steps to move forward.
Incident Response
Point Reliability Droid at production issues and it pulls together the relevant signals—logs, recent deploys, error patterns—to form testable hypotheses. It works through the investigation with specific probes and clear reasoning, moving from symptoms to provide a comprehensive root cause analysis faster.
Available Now
GPT-5 is live as Factory's default model across all droids. Open any session to try it today.