OpenAI released GPT-5.5 on Saturday, positioning the model as a targeted upgrade for enterprise customers who need stronger performance on long-horizon tasks without waiting for the more substantial architectural changes expected in GPT-6. The company describes GPT-5.5 as occupying a specific capability tier — more reliable than GPT-5 on multi-document synthesis and complex instruction-following, but not a wholesale generational leap. It is available immediately to ChatGPT Enterprise and API customers. The release was timed to coincide with an annual developer summit in San Francisco, where OpenAI showcased a series of production automation case studies from companies that have embedded Codex into their engineering workflows.
The Codex platform expansion is in many respects the more strategically significant announcement. OpenAI revealed that Codex now handles over 12 million automated pull requests per month across its enterprise customer base — up from under 3 million in January. The company introduced new fleet management tools that allow engineering managers to assign Codex agents to specific repositories, configure approval workflows, and monitor agent activity through a unified dashboard. Several customers, including a major European bank and two U.S.-based SaaS companies, demonstrated Codex deployments that reduced their QA backlog by over 40 percent. These are not controlled demos: they represent production systems that have been running for multiple months.
Enterprise adoption patterns suggest OpenAI is succeeding at a difficult transition: converting developer curiosity into durable organizational spending. The companies presenting at the summit were not using Codex as a productivity toy but as infrastructure — embedded in CI/CD pipelines, integrated with internal ticketing systems, and monitored by dedicated AI operations teams. This represents a qualitative shift in how enterprises think about AI spend: from experimental line items to capital infrastructure. Competitors including Anthropic and Google are tracking these adoption curves closely, knowing that the enterprises deploying Codex today are unlikely to migrate workflows without significant friction.