OpenAI Launches GPT-5.2 and Codex: Retires GPT-4o and Older Models on February 13

OpenAI has made its biggest model announcement of 2026, introducing GPT-5.2 – the most capable model series yet – while simultaneously announcing the retirement of legacy models from ChatGPT.

GPT-5.2: Three Powerful Variants#

GPT-5.2-Instant#

• Fast, everyday workhorse for work and learning
• Ideal for quick questions, drafting, and general assistance
• Replaces GPT-5 Instant

GPT-5.2-Thinking#

• Built for deeper, more complex work
• Greater polish for coding, summarization, and analysis
• Step-by-step math and logic capabilities
• Planning and decision support

GPT-5.2-Pro#

• Most intelligent and reliable option
• For hard problems where quality matters more than speed
• Available to ChatGPT Pro subscribers

Benchmark Performance#

On GDPval (measuring knowledge work across 44 occupations):

• GPT-5.2 Thinking sets new state-of-the-art score
• Beats or ties top industry professionals on 70.9% of comparisons
• Better at spreadsheets, presentations, code, images, and long contexts

GPT-5.2-Codex: Agentic Coding#

OpenAI also released GPT-5.2-Codex, optimized for autonomous coding:

• Long-horizon work through context compaction
• Stronger cybersecurity capabilities
• GPT-5-Codex-Mini: Cost-effective option with 4x more usage per subscription
• Available in Codex CLI and IDE Extensions

Model Retirements (February 13, 2026)#

OpenAI is retiring these models from ChatGPT:

• GPT-4o – The workhorse model since 2024
• GPT-4.1 and GPT-4.1 mini
• OpenAI o4-mini
• GPT-5 Instant and GPT-5 Thinking

Only 0.1% of users still choose GPT-4o daily. API access remains for developers.

Safety Updates#

GPT-5 Instant is being updated to better recognize and support people in emotional distress, with training guided by mental health experts.

Knowledge Cutoff#

All GPT-5.2 models share a new August 2025 knowledge cutoff.