OpenAI introduces GPT-4.5 AI model with 'higher emotional intelligence'
What's the story
OpenAI has unveiled its latest and most advanced AI model yet, GPT-4.5. Internally, it is code-named 'Orion.'
The new model beats all previous releases in terms of data and computing power used for training.
OpenAI initially said in its white paper that "GPT-4.5 is not a frontier model" and its performance is "below that of o1, o3-mini, and deep research on most preparedness evaluations."
However, hours later, the company released an updated white paper removing that line.
Development process
GPT-4.5 developed using unsupervised learning
The development of GPT-4.5 leveraged a critical technique called unsupervised learning during the "pre-training" phase.
This technique involves massively scaling up the amount of data and computing power used, a strategy that OpenAI has consistently followed in the development of its previous models like GPT-4, GPT-3, GPT-2, and GPT-1.
AI capabilities
GPT-4.5 showcases enhanced world knowledge and emotional intelligence
OpenAI has claimed that the larger size of GPT-4.5 has led to "a deeper world knowledge" and "higher emotional intelligence."
However, the advantages of scaling up data and computing power seem to be plateauing.
On multiple AI benchmarks, GPT-4.5 lags behind newer AI "reasoning" models from Chinese AI company DeepSeek, Anthropic, and even OpenAI itself.
Cost considerations
OpenAI admits high operational costs of GPT-4.5
OpenAI has admitted that GPT-4.5 comes with exorbitant operational costs. The company is evaluating whether to continue offering GPT-4.5 in its API in the long run given these costs.
"We're sharing GPT-4.5 as a research preview to better understand its strengths and limitations," OpenAI said in a blog post, stressing that they're still exploring the model's capabilities and use cases.
Access details
GPT-4.5 to be accessible for ChatGPT Pro subscribers
Beginning Thursday, OpenAI's $200-a-month plan subscribers, ChatGPT Pro, will get access to GPT-4.5 in ChatGPT as part of a research preview.
Developers on paid tiers of OpenAI's API can also start using GPT-4.5 from today onward.
Other users on ChatGPT Plus and ChatGPT Team plans will get access to the model next week, an OpenAI spokesperson confirmed to TechCrunch.
Model comparison
GPT-4.5 not a direct replacement for GPT-4o
OpenAI has clarified GPT-4.5 isn't meant to directly replace GPT-4o, the model powering most of its API and ChatGPT.
While GPT-4.5 supports features like file/image uploads and ChatGPT's canvas tool, it doesn't support certain capabilities like ChatGPT's realistic two-way voice mode.
Despite these limitations, OpenAI notes that on its SimpleQA benchmark, which tests AI models on straightforward factual questions, GPT-4.5 outperforms both GPT-4o and OpenAI's reasoning models in terms of accuracy.