OpenAI has revealed the introduction of GPT-5.2, which is effective right away, with access being implemented in stages beginning Dec. 11.
This represents the second significant upgrade to GPT-5, the much-awaited OpenAI model that first launched on Aug. 7.
In a blog entry unveiling the new collection of ChatGPT models, the organization characterized GPT-5.2 as the “most powerful model series to date for professional knowledge tasks.”
This Tweet is currently not accessible. It may be in the process of loading or has been deleted.
GPT-5.2 consists of multiple models; this series is a component of the broader GPT-5 lineup. Here’s what is included:
GPT-5.2 Instant
GPT-5.2 Thinking
GPT-5.2 Pro
When will you be able to use GPT-5.2?
With a launch date of Dec. 11, GPT-5.2 is currently accessible — although not every user will find it listed as an available model in ChatGPT.
As with prior model releases, OpenAI will initially roll out the model to paying users on Plus, Pro, Go, Business, and Enterprise plans.
For paid users, OpenAI usually releases new models in phases — so if it’s not immediately available through the OpenAI API or ChatGPT, keep checking back.
“We implement GPT-5.2 gradually to ensure ChatGPT remains smooth and dependable,” the organization stated. “If you don’t notice it right away, please try again later. In ChatGPT, GPT-5.1 will still be accessible to paying users for three months as part of legacy models, after which we will retire GPT-5.1.”
If you hold a paid ChatGPT account, visit ChatGPT.com to check if the model is available. Ultimately, all ChatGPT users should gain free access to GPT-5.2, although this access will be restricted for free membership plans.
Performance and benchmarks of GPT-5.2
OpenAI released a new system card indicating that GPT-5.2 shows incremental enhancements on important benchmarks when compared to GPT-5 and 5.1. Additionally, OpenAI noted that GPT-5.2 experiences fewer hallucinations than its predecessors.
GDPval: GPT-5.2 achieved a score of 70.9 percent, as opposed to GPT-5’s 38.8 percent
SWE-Bench Pro: GPT-5.2’s score is 55.6 percent, compared to GPT-5’s 50.8 percent
AIME 2025 math: GPT-5.2 scored 100 percent, while GPT-5 achieved 94 percent