OpenAI rolls out GPT-5.2 with extended Thinking Mode — 1M-token context, parallel chain-of-thought, native tool-call planning shipped to Pro and Plus tiers May 28
OpenAI shipped GPT-5.2 with an extended Thinking Mode on May 28: a 1-million-token context window, parallel chain-of-thought execution, and native tool-call planning. The release matches Anthropic's Claude Opus 4.7 1M-context launch the same day and brings OpenAI's context-window ceiling in line with the Google Gemini 3.1 Ultra and Anthropic frontier tiers. The convergence at 1M+ context across all three major Western labs marks the end of the context-window arms race as a meaningful differentiator.
The 1M-token context window is the substantive capability piece. GPT-5.0 shipped in early 2026 with a 256K context ceiling and a separate o-series Thinking Mode line. GPT-5.1 unified the lines in late Q1 with a 512K ceiling. GPT-5.2 doubles the ceiling to 1M and integrates Thinking Mode as a Pro-tier default rather than an opt-in flag. The architectural commitments — parallel chain-of-thought during the planning phase, native tool-call orchestration during execution, and the per-tier compute-allocation tiers that govern Thinking-Mode depth — are the operational shape of the release. For developers, the practical change is that Pro-tier API calls can target the full 1M context without manual chunking, and the Thinking-Mode opt-in flag is no longer required for the depth-of-reasoning workloads where it was previously needed.
The competitive convergence is what makes the moment structurally meaningful. Anthropic's Claude Opus 4.7 launched the same day with a 1M-token context, and Google's Gemini 3.1 Ultra at 2M extends the ceiling further. With all three major Western frontier labs at 1M-or-above context, the context-window axis is no longer a meaningful differentiator at the top tier. The competitive frame shifts to execution architecture (Thinking Mode depth, parallel tool-call orchestration, persistent-agent integration) and ecosystem depth. OpenAI's ChatGPT Agent Mode priority tier launched the same day pairs the GPT-5.2 model upgrade with the execution-architecture upgrade, which is the procurement-relevant story.
OpenAI — GPT-5.2 Thinking Mode launch May 28 2026 1M context → · TechCrunch — OpenAI GPT-5.2 release 1M-token context Pro Plus tiers → · The Verge — GPT-5.2 Thinking Mode parallel chain of thought →