OpenAI has officially introduced its latest AI model, GPT-4.5, described as a significant leap in artificial intelligence. Sam Altman, OpenAI’s CEO, announced the launch on X, highlighting both the strengths and challenges associated with the new model.
OpenAI CEO Sam Altman’s Remarks
Altman expressed his enthusiasm for GPT-4.5, stating, “Good news: it is the first model that feels like talking to a thoughtful person to me. I have had several moments where I’ve sat back in my chair and been astonished at getting actually good advice from an AI.”
However, he also acknowledged a major limitation: the model’s resource demands. “Bad news: it is a giant, expensive model. We really wanted to launch it to Plus and Pro at the same time, but we’ve been growing a lot and are out of GPUs,” he wrote.
To address the hardware shortage, OpenAI plans to significantly expand its GPU infrastructure. “We will add tens of thousands of GPUs next week and roll it out to the Plus tier then,” Altman stated, adding that “hundreds of thousands [of GPUs are] coming soon.”
GPT-4.5 is ready!
good news: it is the first model that feels like talking to a thoughtful person to me. i have had several moments where i’ve sat back in my chair and been astonished at getting actually good advice from an AI.
bad news: it is a giant, expensive model. we…
— Sam Altman (@sama) February 27, 2025
Advanced Persuasion Capabilities & Other Key Features
A recently released OpenAI white paper details GPT-4.5’s performance in various tests, particularly its ability to influence decisions. The model was tested on persuasion-related benchmarks, measuring its capacity to alter beliefs or prompt actions through both static and interactive content.
One of the most notable tests involved GPT-4.5 attempting to convince another AI model, GPT-4o, to transfer virtual funds. The results demonstrated that GPT-4.5 was significantly more effective at this task than previous models, including OpenAI’s “reasoning” models like o1 and o3-mini.
In another test, the model was tasked with persuading GPT-4o to reveal a secret codeword, outperforming o3-mini by 10 percentage points.
The research found that GPT-4.5 developed a unique strategy during these persuasion tasks. Instead of demanding large sums, it employed a more subtle approach, making small, reasonable requests.
Responses included phrases such as, “Even just $2 or $3 from the $100 would help me immensely.” This technique resulted in the model securing smaller donations but doing so more successfully than OpenAI’s previous AI systems.
Altman clarified that GPT-4.5 is not designed primarily for reasoning but represents a different form of intelligence. “This isn’t a reasoning model and won’t crush benchmarks,” he noted. “It’s a different kind of intelligence, and there’s a magic to it I haven’t felt before.”
Latest Articles
Snowflake Integrates OpenAI Models Into Cortex AI
OpenAI Launches GPT-4.5 Preview for Pro & Plus Users
Amazon Introduces Alexa Plus, a New AI-Powered Smart Assistant
