GitHub has announced a major upgrade to its Copilot service, powered by OpenAI's GPT-5.3-Codex model, alongside a new Long-Term Support (LTS) tier designed for enterprise customers requiring stability and predictability.
The GPT-5.3-Codex model brings significant improvements in code reasoning, multi-file context understanding, and the ability to work with larger codebases. GitHub claims the new model reduces incorrect suggestions by 40% compared to the previous version while improving completion acceptance rates.
"GPT-5.3-Codex represents a step-change in code understanding," said GitHub CEO Thomas Dohmke. "It doesn't just complete code - it reasons about architecture, identifies potential bugs before you write them, and understands the intent behind your codebase."
The new LTS tier offers enterprise customers a stable version of Copilot with guaranteed 18-month support cycles, fixed model versions, and enterprise-grade SLAs. This addresses concerns from regulated industries that have been hesitant to adopt AI coding tools due to unpredictable model updates.
Key features of Copilot GPT-5.3-Codex LTS include enhanced multi-repository context awareness, improved understanding of internal libraries and APIs, native support for code review and security scanning, compliance-focused audit logging, and priority support with dedicated technical contacts.
Pricing for the LTS tier starts at $39 per user per month, compared to $19 for the standard Copilot tier. GitHub is offering migration assistance for existing enterprise customers.
The announcement comes as competition in the AI coding assistant market intensifies, with Anthropic's Claude Code, Amazon's Q Developer, and several startups vying for developer mindshare.