We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Nov 19·5 min read
Building more with GPT-5.1-Codex-Max
TL;DR
Previous coding models struggled with long-running, project-scale development tasks requiring sustained reasoning across multiple files and dependencies. OpenAI built GPT-5.1-Codex-Max, an agentic coding model (one that can plan and execute multi-step tasks autonomously) with improved reasoning and token efficiency.
✦ Why It Matters
Engineers can now tackle multi-file projects and complex architectures with AI assistance that sustains reasoning across entire workflows.
Key Takeaways
- Use GPT-5.1-Codex-Max for multi-file projects requiring sustained reasoning across dependencies.
- Leverage improved token efficiency to reduce computational costs on large codebases.
- Deploy agentic capabilities to automate multi-step coding tasks without manual intervention.
- Adopt the model for long-running development workflows exceeding single-file scope.
Full Summary
Existing code generation models like earlier Codex versions faced limitations when handling large, multi-file projects requiring sustained reasoning across dependencies and architectural decisions. OpenAI released GPT-5.1-Codex-Max, an agentic coding model designed to autonomously plan and execute multi-step coding tasks at project scale.
The model incorporates enhanced reasoning capabilities—the ability to break down complex problems logically—and improved token efficiency, meaning it produces better outputs while consuming fewer computational tokens (units of text processed). Built for long-running development workflows, GPT-5.1-Codex-Max targets scenarios where engineers need sustained code generation across interconnected components.
The model demonstrates faster inference speed and higher code quality on extended tasks compared to prior versions, enabling engineers to tackle larger codebases with fewer manual interventions.