GPT-5.2 is rolling out to ChatGPT: How to try it

OpenAI has announced its launching GPT-5.2 effective immediately, with access rolling out in phases starting Dec. 11.

This is the second major update to GPT-5, the long-awaited OpenAI model that launched on Aug. 7.

In a blog post announcing the new series of ChatGPT models, the company called GPT-5.2 the "most capable model series yet for professional knowledge work."

This Tweet is currently unavailable. It might be loading or has been removed.

GPT-5.2 is actually a series of models; the series itself is part of the larger GPT-5 series. Here's what we're looking at:

• GPT-5.2 Instant

• GPT-5.2 Thinking

• GPT-5.2 Pro

When can you start using GPT-5.2?

With a Dec. 11 release date, GPT-5.2 is available now — though not all users will see it listed as an available model in ChatGPT.

As with previous model rollouts, OpenAI will first deploy the model to paid users on Plus, Pro, Go, Business, and Enterprise plans.

Even for paying customers, OpenAI tends to make its new models available in phases — so if it's not available immediately via the OpenAI API or ChatGPT, keep checking.

"We deploy GPT‑5.2 gradually to keep ChatGPT as smooth and reliable as we can," the company wrote. "If you don’t see it at first, please try again later. 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."

If you have a paid ChatGPT account, go to ChatGPT.com to see if the model is available. Eventually, all ChatGPT users should have free access to GPT-5.2, though that access will be limited on free membership plans.

GPT-5.2 benchmarks and performance

OpenAI published a new system card that shows GPT-5.2 makes iterative improvements on key benchmarks compared to GPT-5 and 5.1. In addition, OpenAI said that GPT-5.2 hallucinates less than previous models.

• GDPval: GPT-5.2 scored 70.9 percent, while GPT-5 scored 38.8 percent

• SWE-Bench Pro: GPT-5.2 scored 55.6 percent, while GPT-5 scored 50.8 percent

• AIME 2025 math: GPT-5.2 scored 100 percent, while GPT-5 scored 94 percent

• FrontierMath (Tier 1–3): GPT-5.2 scored 40.3 percent, while GPT-5 scored 31.0 percent

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.