OpenAI announced today that it's launching GPT-5.2, the newest model in its GPT-5 series. The new model will start rolling out immediately, with paid ChatGPT customers getting access first.
In a blog post announcing the new model — which is actually a series of models, comprised of GPT‑5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro — OpenAI said that GPT-5.2 makes noticeable improvements in math and science, imaging, coding, handling agentic tasks, and overall accuracy. The company called GPT-5.2 its "most capable model series yet for professional knowledge work."
The new model comes at a difficult time for OpenAI, which is rumored to be in a "code red" state over stronger competition from rivals like Google Gemini and spreading fears of an AI bubble.
Ever since it launched ChatGPT in 2022, OpenAI has been securely on top of the AI industry. However, the company is in an increasingly precarious position. Google has an almost unfathomable amount of training data at its disposal, and Google AI products like Gemini 3, Veo 3, and Nano Banana have outperformed GPT-5, the new model OpenAI launched earlier this year, in many respects.
Still, ChatGPT is by far the most popular AI chatbot in the world, with an estimated 700 million weekly active users.
How to try GPT-5.2
The new GPT-5.2 models will start rolling out immediately, though access may not be available right away to all users. As per usual, OpenAI will launch the models to paid users on the Plus, Pro, Go, Business, and Enterprise accounts.
As of this writing, GPT-5.2 was not yet available for this reporter, and the rollout will likely happen in phases.
"We deploy GPT‑5.2 gradually to keep ChatGPT as smooth and reliable as we can; if you don’t see it at first, please try again later," OpenAI wrote in a blog post. "In ChatGPT, GPT‑5.1 will still be available to paid users for three months under legacy models, after which we will sunset GPT‑5.1."
OpenAI says GPT-5.2 makes key improvements in safety, accuracy, and performance benchmarks
The AI industry relies on standardized benchmark tests to demonstrate how well models perform, and companies like OpenAI also have their own internal tests. In addition, AI leaderboards like LMArena let users compare and rank various AI models. While GPT-5.2 has already appeared near the top of LMArena's AI coding leaderboard, it will take more time to see how users rate the new series of models against the competition. However, OpenAI released a new model card for GPT-5.2 on Dec. 11, which shows that the model makes across-the-board improvements in a variety of areas, which isn't surprising.
Most notably, OpenAI says that GPT-5.2 is more accurate and will produce fewer hallucinations compared to GPT-5.1. OpenAI's documentation states that GPT-5.2 Thinking has an average hallucination rate of 10.9 percent, compared to 16.8 percent and 12.7 percent for GPT-5 Thinking and GPT-5.1 Thinking, respectively. When GPT-5.2 is given access to the web via a browser, its hallucination rate drops to 5.8 percent.
In its blog post, OpenAI also states that GPT-5.2 scores more highly on benchmark tests for coding, science and math, performing economically valuable tasks, computer vision, and agentic work involving third-party tools. OpenAI also highlighted GPT-5.2's improved abilities with spreadsheets, in particular.
OpenAI says GPT-5.2 is safer for users with mental health problems
Lately, OpenAI has been accused of endangering ChatGPT users with mental health issues. Due to well-documented sycophancy problems, ChatGPT reportedly encouraged delusions and conspiratorial thinking on some users, who later died by suicide. OpenAI is now facing wrongful death suits, including a new suit that was just revealed for the first time today by the Wall Street Journal, in which a ChatGPT user killed himself shortly after killing his own mother.
OpenAI says that according to its internal tests, GPT-5.2 has a better response to users with mental health problems.
"With this release, we continued our work to strengthen our models’ responses in sensitive conversations, with meaningful improvements in how they respond to prompts indicating signs of suicide or self harm, mental health distress, or emotional reliance on the model. These targeted interventions have resulted in fewer undesirable responses in both GPT‑5.2 Instant and GPT‑5.2 Thinking as compared to GPT‑5.1 and GPT‑5 Instant and Thinking models."
Mashable has not been able to independently verify these results, and the GPT-5.2 system card has scant details on how safety performance was measured in this context.
For more information, check out the OpenAI blog post announcing GPT-5.2 or read the new GPT-5.2 system card.
If you're feeling suicidal or experiencing a mental health crisis, please talk to somebody. You can call or text the 988 Suicide & Crisis Lifeline at 988, or chat at 988lifeline.org. You can reach the Trans Lifeline by calling 877-565-8860 or the Trevor Project at 866-488-7386. Text "START" to Crisis Text Line at 741-741. Contact the NAMI HelpLine at 1-800-950-NAMI, Monday through Friday from 10:00 a.m. – 10:00 p.m. ET, or email info@nami.org. If you don't like the phone, consider using the 988 Suicide and Crisis Lifeline Chat. Here is a list of international resources.
Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
from Mashable https://ift.tt/SeQk6Xy
via IFTTT
Comments
Post a Comment