OpenAI Unveils GPT-5.2 Amid Intensifying Competition With Google

• OpenAI has launched GPT-5.2 for developers and professional use.
• The release follows reports of “code red” internally due to pressure from Google.
• GPT-5.2 emphasizes reasoning, programming, and enterprise scenarios.

---

OpenAI has announced the release of its new flagship GPT-5.2. The firm is positioning the model as the most advanced version of AI for professional and enterprise use.

GPT-5.2 is available to paid ChatGPT users and developers via API in three variants:

• Instant version is focused on quick everyday tasks like searching for information and writing texts;
• Thinking mode is designed for complex structured work, including programming, document analysis and mathematical calculations;
• Pro mode is designed to maximize accuracy and reliability for mission-critical tasks.

OpenAI chief product officer Fidji Simo said the model was created with practical value in mind. She said GPT-5.2 is better at creating spreadsheets and presentations, writing code, analyzing images, working with context and managing multistep projects.

The model’s release comes amid the company’s product’s rivalry with Google’s Gemini 3 Pro. The latter leads the LMArena rankings in most metrics except programming.

Earlier, it was reported that OpenAI CEO Sam Altman announced a “code red” mode to improve the company’s AI models under development. This is allegedly due to declining ChatGPT traffic and concerns about losing market share to Google.

Some analysts claim that there were discussions internally about delaying the launch of GPT-5.2 for additional refinement. However, OpenAI has bet on an accelerated release and expanding its presence in the enterprise segment.

The firm claims that GPT-5.2 shows improved performance in programming, math, science, computer vision and long context work. According to internal company estimates, the Thinking version outperforms the Gemini 3 Pro and Claude Opus 4.5 in a number of logical thinking tests, including software development tasks and scientific benchmarks.

Results on some benchmarks for GPT-5.2. Data: OpenAI.

At the same time, reliance on reasoning models increases computational costs, experts say. According to media reports, OpenAI is already spending more on inferencing than it previously claimed, paying for computing resources directly rather than through subsidized access to cloud infrastructure.

The company expects to compensate for rising costs by scaling products and increasing revenue, experts believe. OpenAI representatives emphasized that the efficiency of their models continues to grow over time.

Recall, OpenAI warned that the capabilities of its advanced models are developing faster than expected. According to the developers, future generations of AI may get a “high” level of cybersecurity risk.