OpenAI released GPT-5.5 and GPT-5.5 Pro to the Chat Completions, Responses, and Batch APIs in April 2026, introducing a 1M token context window, native tool search, built-in computer use, Skills, MCP, hosted shell, apply patch, and default medium reasoning effort (https://developers.openai.com/api/docs/changelog).

March 2026's GPT-5.4 rollout first added computer use and native compaction; the 5.5 release integrates these into a single frontier model while adding MCP and Skills layers absent from prior coverage, directly extending patterns seen in the o1 reasoning series and the 2024 Operator prototype (OpenAI o1 System Card, 2024; OpenAI DevDay 2024 announcements). Original changelog omits explicit linkage to the updated Agents SDK sandboxing and memory controls released simultaneously, which close the loop for production agent deployment.

Synthesis of the OpenAI changelog, Anthropic's October 2024 Claude 3.5 Sonnet computer-use benchmark paper, and SemiAnalysis's Q1 2026 inference scaling report shows GPT-5.5 Pro's variable compute tier for "tougher problems" mirrors the o1 Pro pricing precedent and will compress competitor margins; 50% Batch discounts on GPT Image 2 further signal deliberate price-leadership tactics not highlighted in initial reporting.