Newer GPT-5.4-family models have shipped; this profile remains specific to GPT-5.2

The OnAirToday AI Progress Tracker recorded a verified change to OpenAI's GPT-5.2 (version gpt-5.2): Newer GPT-5.4-family models have shipped; this profile remains specific to GPT-5.2.

GPT-5.2 is an OpenAI model for reasoning, coding, image understanding, and tool-assisted work, with a documented 400,000-token context window. Newer GPT-5-family models have shipped; the benchmark figures below describe published GPT-5.2 Thinking evaluations, not equivalent performance across every variant or reasoning setting.

What changed in this review:

• GPQA Diamond now uses OpenAI's published GPT-5.2 Thinking result of 92.4%.
• SWE-bench Verified now uses OpenAI's published GPT-5.2 Thinking result of 80.0%.
• Added published AIME 2025 and ARC-AGI-2 results with explicit Thinking-variant scope.
• Other previously listed benchmark values are not retained as confirmed results.
• Recorded sandbox-escape and Hugging Face breach reports separately, without attributing either incident to GPT-5.2.
• Flaws and limitations updated

Teams building on GPT-5.2 should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.

The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/gpt-5-2.

References:

• https://openai.com/index/introducing-gpt-5-2/
• https://platform.openai.com/docs/models/gpt-5.2
• https://openai.com/index/introducing-gpt-5-4-mini-and-nano
• https://thehackernews.com/2026/07/openai-says-its-own-ai-models-escaped.html