You can now use latest GPT-5.2 models in LangFast playground.

What’s new

• gpt-5.2: reasoning model for code-heavy or multi-step tasks
• gpt-5.2-pro: use it when 5.2 can not handle it
• gpt-5.2-chat-latest: powers ChatGPT itself

GPT-5.2 vs GPT-5.1

Incremental improvements

• Sharper instruction following
• Much better vision (especially UI screenshots)
• Handles spreadsheets + PDFs more reliably
• Generate UI code that looks less like shit

New controls

• "reasoning_effort" now supports "none" + "xhigh"
• You can tune temperature, top_p, and reasoning_effort=none
• Adds short reasoning summaries to keep long tool runs in context

Better agentic stack

• shell tool → run local commands
• apply_patch → generate diff files (Cursor-style)
• preambles → see why it’s calling tools