OpenAI launched GPT-4.5 a new AI model, which the company says is emotionally intelligent. It also claims to be less susceptible to hallucinations.
OpenAI has now released the final version of its pretraining model. It is available for ChatGPT users to purchase at $200 per monthly.
This release has ignited debates about whether AI leader, Anthropic is losing out to competitors such as DeepSeek and Anthropic who are introducing models with better reasoning abilities.
GPT-4.5 costs 15-30 times more to run than GPT-4o.
OpenAI, despite its improved claims, has been cautious. CEO Sam Altman has downplayed expectations, stating that GPT 4.5 is not “a frontier model.”
A last-minute change was made to the company’s technical paper. The statement that GPT-4.5 is not a sophisticated AI system has been removed, which raised more speculation.
Anthropic’s Claude 3.7 Sonnet, and China’s DeepSeek R1 have gained traction in recent years, both of which offer more complex reasoning abilities.
OpenAI’s dominant position in the AI race is less sure as the AI race continues to intensify, particularly with GPT-5 on the horizon.
Benchmark data raises doubts
Fortune reported that although OpenAI’s latest model has outperformed GPT-4o in several key metrics it failed to achieve a breakthrough when it comes to logical reasoning and multilingual problem solving.
Initial benchmark comparisons show that GPT-4.5 is unable to compete with Anthropic’s latest Claude, a model that uses a combination of intuitive responses and deep reasoning.
Unlike GPT 4.5, Claude 3.7 Sonnet makes a decision in real-time whether it will generate immediate responses or use a “chain of thoughts” to fine tune its answers.
This makes Claude more flexible to different queries. OpenAI’s newest release does not have this level of flexibility. This raises concerns about the AI models becoming outdated as markets evolve.
The social media reaction has been muted, and some AI researchers have posted benchmark results that are not impressive.
AI expert Gary Marcus called GPT-4.5 “a nothingburger”, reflecting the growing doubts that OpenAI can maintain its technological advantage.
OpenAI shifts AI Strategy
The GPT-4.5, also known as Orion, is the final model to have been built with OpenAI’s pretraining strategy. This relied heavily on increasing model size and input data.
OpenAI will shift its focus to reasoning models that use reinforcement learning in order to improve logical processing during testing.
Anthropic, Google and other companies invest heavily in computational models which can adjust resources dynamically based on the complexity of a task.
DeepSeek is a Chinese AI company that has introduced models with reasoning to challenge OpenAI’s technology.
OpenAI faces increasing pressure from the competition to provide a model that is truly next generation.
Sam Altman, CEO of Sam Altman Inc. has announced that GPT-5 is coming in the next few months. It promises a hybrid method that blends GPT style fluency and the logic-by-step reasoning models.
It remains to be determined if this change will allow OpenAI to regain its lead.
OpenAI, DeepSeek and Anthropic face off in DeepSeek’s challenge
OpenAI is not the only AI player in this landscape. There are many others who have emerged to challenge OpenAI.
Anthropic is a leading AI in the reasoning field, while DeepSeek’s R1 has shown strong results for coding and mathematic reasoning.
Meta and Google continue to improve their AI products, using vast computing resources to advance the limits of generative AI.
OpenAI’s technical supremacy is now under question, and the AI industry has entered a new phase in which no one company can claim to have a clear advantage.
OpenAI must prove that, as GPT-5 launches near, it is able to keep pace with an industry moving towards models based on reasoning.
Artificial intelligence will be defined by the companies who adapt to the new realities.
The post OpenAI GPT-4.5 launches signals a shift in the AI race, as Anthropic and DeepSeek gain more ground could be updated as new updates occur.
