OpenAI has released GPT-5.2, its most advanced AI model to date, arriving just one week after CEO Sam Altman declared a "code red" to refocus the company on competing with Google's Gemini and Anthropic's Claude.
Three Model Variants
GPT-5.2 comes in three flavors:
- •Instant - Speed-optimized for routine queries like information-seeking, writing, and translation
- •Thinking - Excels at complex structured work: coding, document analysis, math, and planning
- •Pro - Maximum accuracy and reliability for difficult problems
Key Improvements
The new models deliver significant improvements over GPT-5.1:
- •38% fewer errors in Thinking model responses
- •30% reduction in hallucinations compared to GPT-5.1 Thinking
- •Improved long context handling across hundreds of thousands of tokens
- •Better at creating spreadsheets, presentations, and multi-step projects
Technical Specifications
Spec | Value |
|---|---|
| Context Window | 400,000 tokens |
| Max Output | 128,000 tokens |
| Knowledge Cutoff | August 31, 2025 |
| Pricing | $1.75/M input, $14/M output |
4 items
The pricing represents a 1.4x increase from GPT-5.1.
Benchmark Performance
On OpenAI's GDPval benchmark measuring knowledge work across 44 occupations, GPT-5.2 Thinking sets a new state-of-the-art score, beating or tying top industry professionals on 70.9% of comparisons.
The model also outperforms Gemini 3 and Claude Opus 4.5 on:
- •SWE-Bench Pro (real-world software engineering)
- •GPQA Diamond (doctoral-level science)
- •ARC-AGI suites (abstract reasoning)
Availability
GPT-5.2 is rolling out to ChatGPT paid users and developers via the API starting today.
Source: OpenAI Blog