OpenAI Unveils GPT-4.5 Orion: What You Need to Know

OpenAI has officially launched its latest AI model, GPT-4.5, also known as Orion. This new model builds on its predecessors with enhanced computing power and a vast dataset. While it marks another step forward in AI development, OpenAI has clarified that GPT-4.5 is not considered a “frontier” AI breakthrough, indicating that it is an iterative improvement rather than a groundbreaking leap.

Who Can Access GPT-4.5?

Currently, GPT-4.5 is available exclusively to ChatGPT Pro subscribers, who pay $200 per month for premium access. Developers using OpenAI’s paid API can also start testing the model immediately. OpenAI has announced that ChatGPT Plus and ChatGPT Team users will gain access next week, expanding the model’s reach to a broader audience.

How Does GPT-4.5 Compare to Previous Models?

OpenAI has consistently improved its AI models by scaling up computing power and training data. The company followed this same strategy with Orion, expecting notable advancements. However, early performance benchmarks show a more nuanced picture.

Where GPT-4.5 Excels

• Improved General Knowledge: OpenAI claims that GPT-4.5 has deeper world knowledge, meaning it can generate more contextually accurate responses.
• Better Emotional Intelligence: The model reportedly understands human emotions and intent more effectively, leading to more natural and empathetic interactions.
• Reduced Hallucinations: One of the key improvements in GPT-4.5 is its lower tendency to generate false or misleading information.

Where GPT-4.5 Falls Short

• Struggles with Advanced Reasoning: Despite improvements in general knowledge, GPT-4.5 underperforms in complex reasoning tasks compared to competitors like DeepSeek’s R1 and Anthropic’s Claude 3.7 Sonnet.
• Not the Most Powerful Model: Some AI benchmarks indicate that other models perform better in areas like problem-solving and logic-based reasoning, raising questions about whether OpenAI’s scaling approach is still the best strategy.

Is GPT-4.5 Worth the Price?

One major concern about GPT-4.5 is its high cost. OpenAI is charging:

• $75 per million input tokens (about 750,000 words)
• $150 per million output tokens

For comparison, GPT-4o, OpenAI’s widely used AI model, costs just $2.50 per million input tokens and $10 per million output tokens. Given the price difference, some developers may hesitate to switch to GPT-4.5 unless they see a clear advantage in performance. OpenAI is also evaluating whether to continue offering API access to GPT-4.5 in the long run.

Can GPT-4.5 Understand Humans Better?

One area where OpenAI claims a major improvement is in understanding human emotions and social interactions. In an experiment, GPT-4.5 was compared to GPT-4o and o3-mini in two tasks:

1. Generating an SVG image of a unicorn – GPT-4.5 was the only model that successfully produced a recognizable unicorn.
2. Responding to a user struggling with a failed test – GPT-4.5’s reply was deemed the most socially appropriate and empathetic, showing its enhanced emotional intelligence.

The Bigger Picture: Is AI Scaling Losing Steam?

The release of GPT-4.5 has reignited a debate in the AI community: Are we reaching the limits of traditional AI scaling?

Ilya Sutskever, OpenAI’s co-founder, has previously suggested that simply increasing computing power and data may no longer deliver significant performance gains. This has led AI companies, including OpenAI, to shift their focus toward reasoning models, which prioritize deeper problem-solving over sheer processing power.

OpenAI has already announced plans to merge the GPT series with its “o” reasoning models, starting with GPT-5 later this year. This could mark a shift in AI development strategies, moving away from brute-force improvements toward more efficient and intelligent problem-solving techniques.

What’s Next for OpenAI?

While GPT-4.5 may not be a revolutionary AI model, it serves as an important stepping stone toward the future of AI. OpenAI’s focus now appears to be on combining generative capabilities with stronger reasoning skills, potentially paving the way for more advanced AI systems in the coming years.

We will soon see whether this strategy leads to the next big AI breakthrough. For now, GPT-4.5 offers an intriguing glimpse into what’s next in AI evolution.

See Also: AI Benchmark Controversy: OpenAI and xAI Clash Over Grok 3’s Performance Claims