OpenAI announced GPT-5.2 on December 11, 2025, claiming improved performance on coding, spreadsheet creation, image perception, and long-context understanding, with three tier variants (Instant, Thinking, Pro). The model reportedly tops SWE-Bench Pro (agentic coding) and GPQA Diamond benchmarks, and scores 70.9% vs. top professionals on OpenAI's own GDPval evaluation. CEO Altman stated Google's Gemini 3 release had less competitive impact than feared and expects OpenAI to exit "code red" status by January. This signals continued gen-over-gen capability advancement and the company's ability to release high-capability models on rapid cadence, though benchmark comparisons show Anthropic's Opus 4.5 higher on an alternative coding test (SWE-Bench Verified).
183 other tracked developments touching the same claims.