The rumors were true, and "Code Red" It’s Out: OpenEye today announced the release of its new Frontier Large Language Model (LLM) family: GPT-5.2.
It comes at a critical moment for the AI ​​pioneer, which has faced pressure after rival Google’s Gemini 3 LLM took over the top spot on major third-party performance leaderboards and several key benchmarks, although Openei leaders stressed in a press briefing that the timing of the release had been discussed and worked on.
OpenAI defines GPPT-5.2 as "Still the most capable model series for professional knowledge work," Reclaiming the performance crown with significant gains in reasoning, coding, and agent workflows.
"This is our latest frontier model and the strongest ever on the market for professional use," Openei Applications CEO Fidji Simu said during a press briefing today. "We designed 5.2 to unlock more economic value for people. It is better at creating spreadsheets, presenting presentations, writing code, understanding images, understanding longer contexts, using tools, and managing complex, multi-faceted projects."
GPT-5.2 includes a massive 400,000-token context window—which requires hundreds of documents or large code repositories at once.
The model also features a knowledge cutoff of August 31, 2025, ensuring that it is up-to-date with relatively recent world events and technical documentation. This clearly includes "Rationale of token support," Verifying the basic architecture uses the thinking process of China that is popularized by it "O1" Series
‘Code Red’ reality check
It has come after release InformationReport of an emergency "Code Red" Instruction to Openai staff from CEO Sam Altman. "Quality difference" Exposed by Gemini 3. Verge Similarly, the release time of GPT 5.2 has also been reported before the official announcement.
During the briefing, Openei executives acknowledged the directive but pushed back on the narrative that the model was delivered solely to answer Google.
"It is important to note that this has been in the works for many, many months." Simo told reporters. He clarified that while "Code Red" Helped focus the company, it wasn’t the only driver of the timeline.
"We’ve announced this code to really signal to the company that we want to marshal resources in a certain area … but that’s not why it’s coming out specifically this week."
Team lead Max Schwarzer echoed the sentiment after Opanai’s training to dispel the idea of ​​a nervous start. "We have been planning for this release for a very long time … we talked about this particular week several months ago."
Under the hood: Quick, thoughtful and supportive
OpenAI is splitting the GPT-5.2 release into three separate tiers within ChatGPT, possibly a strategy designed to balance its massive compute costs. "Reasoning" Model with user demand for speed:
GPT-5.2 Quick: Optimized for speed and everyday tasks like writing, translating, and searching for information.
GPT-5.2 Thinking: designed for "Complex, structured work" And long-running agents, this model leverages deep reasoning chains to handle coding, math, and multitasking projects.
GPT-5.2 Pro: The new heavyweight champion. Openei describes it as: "A smart and highly reliable option," Delivering the highest accuracy for difficult queries where quality outweighs latency.
For developers, models are readily available in ASPI gpt-5.2for , for , for , . gpt-5.2-chat-latest (immediate), and gpt-5.2-pro.
Number: Beating the benchmark
Unlike previous launches that often focused on creativity or "Vibes," This release is all about hard metrics – especially those that target it "Professional knowledge work" The gap where competitors have recently gained ground.
Openei highlighted a new benchmark called GDPval, which measures performance "Well-specified knowledge tasks" In 44 professions.
"According to expert human judges, the GPT-5.2’s thinking is now state-of-the-art on this benchmark … and is beating top industry professionals by 70.9% on well-specified professional tasks such as spreadsheets, presentations, and document creation." Simo said.
In the critical field of coding, OpenAI is claiming a decisive lead. Schwarzer noted that on SWE Bench Pro, a rigorous assessment of real-world software engineering, the GPT-5.2 Souch set a new state-of-the-art score of 55.6 percent.
He emphasized that this is the benchmark "More pollution resistant, challenging, diverse and industrially relevant than previous benchmarks such as SWE Bench Certification."Other key benchmark results include:
GPQA Diamond (Science): The GPT-5.2 Pro scored 93.2%, besting the GPT-5.2 Think (92.4%) and surpassing the GPT 5.1 Think (88.1%).
Frontiermoth: On Levels 1-3 problems, the GPT-5.2 thought solved 40.3%, a significant jump from the 31.0% achieved by its predecessor.
Arc-Eg-1: The GPT-5.2 Pro is reportedly the first model to surpass the 90 percent threshold on this general reasoning benchmark, scoring 90.5%
The price of intelligence
Performance comes at a premium. While ChatGPT subscription prices remain unchanged, API costs for the new flagship models are steep compared to previous generations, reflecting higher compute demands. "to think" Models.
GPT-5.2 (Thinking): At a price 75 1.75 per 1 million input tokens and $14 per 1 million output tokens.
GPT-5.2 Pro: The cost increased significantly $21 per 1 million input tokens and 8 168 per 1 million output tokens.
Openei argues that despite the high price, the model of the model "More token efficiency" And the ability to solve tasks in fewer turns makes it economically viable for high-value enterprise workflows.
Image Generation: Nothing New Yet … But ‘More to Come’
During the briefing, VentureBeat asked OpenEye attendees if the new release included any enhancements to image generation capabilities, noting the excitement surrounding similar features in recent rival launches such as Google’s Gemini 3 Image aka Nano Banana Pro.
Unfortunately for those looking to replicate the text and info-type heavy graphics and image editing capabilities, OpenAI executives clarified that GPT-5.2 is not up to date with current image improvements over GPT-5.1 and OpenAI’s integrated DEL-E3 and GPT-4O native image generation models.
"At Image General, nothing to announce today, but more to come." Simo said. He recognized the popularity of this feature, and added, "We know it’s a very important use case that people love, that we introduced to the market, and so there’s definitely more to come."
Openii’s head of training Aidan Clarke also declined to comment on the specifics of the visual generation, saying only, "I can’t really speak to Image General myself."
The ‘Mega Agent’ era
Beyond raw scores, OpenAI is positioning GPT 5.2 as the engine for a new generation of gamers. "Long acting agents" Able to execute multi-step workflows without human hand preparation."
Box found that 5.2 could extract information from long, complex documents about 40% faster, and it saw a 40% increase in reasoning accuracy for the life sciences and healthcare." Simo said.
He also noted that the idea informed the model "outperforms 5.1 across every dimension… and it excels at the kind of really ambiguous, longer rising tasks that define real knowledge work."Coding startups like Augment Code found the model, Schwarzer added "Delivering significantly stronger deep code capabilities than any previous model," This is why it was chosen to power their new code review agent. Various capabilities have also seen an upgrade.
A new evaluation called Screenshot Pro, which tests a model’s ability to understand GUI screenshots, shows GPT-5.2 achieving 86.3% accuracy, compared to just 64.2% for GPT-5.2.
Science and reliability
OpenAI leaders also emphasized the utility of the model for scientific research, and sought to expand the conversation beyond simple chatbots.
Training team lead Aidan Clark shared an example of a senior immunology researcher testing the model.
"They experimented with this to raise very important answerable questions about the immune system, saying," Clarke said. "This immunology researcher reported that GPT-5.2 offered sharper questions and stronger explanations and why those questions … are more important than any previous pro-model.
"Reliability was another key focus. Schwarzer claimed the new model "GPT-5.1 to substantially halchinites," Noting that on a set of identified questions, "Answer errors are 38% less frequent."
The ‘vibe’ shift
Interestingly, Openei admitted that not every customer immediately prefers the new models.
When asked why legacy models like the GPT 5.1 would continue to be available, Schwarzer admitted: "Models change a little each time.
"Some users may find that they prefer the vibration of the previous model, although we think the latest board is generally much better at it." Schwarzer said. He also noted that some are for enterprise users "Indeed indicated for a specific model," may be "small pressure," Access to older versions is required.
Safety, ‘mature mode’, and the future roadmap
Addressing security concerns, Simo confirmed that the company is preparing to roll out one "Adult mode" In the first quarter of next year, after the implementation of the new age prediction system.
"We are in the process of improving it." Simo said about age prediction technology.
"We want to do this before launching adult mode."Looking further ahead, industry reports suggest that OpenAI is working on a more fundamental architectural change under the code name "Project Garlic," Targeting a flagship release in early 2026.
While executives did not comment on a specific future roadmap during the briefing, Simo remained optimistic about the economics of its current trajectory.
"If you look at historical trends, compute has grown by about 3x every year for the last three years," He explained. "Revenues have also increased at the same pace…creating this virtuous cycle."
Clark added that efficiency is improving rapidly: "The model we’re releasing today scores even better (on Arc-AGI) with about 400 times less cost and associated less compute." Compared to a year ago models.
GPT-5.2 Instant, Thought, and Pro Chat begin rolling out today to paying users (Plus, Pro, Team, and Enterprise) in GPT. The company notes that the rollout will be gradual to maintain stability.