OpenAI unveils GPT 4.5 Orion
OpenAI has unveiled its latest AI language model, GPT-4.5, internally referred to as ‘Orion’. This model represents the company’s most advanced and largest AI system to date, developed with increased computational resources and data compared to its predecessors. Despite its enhanced capabilities, OpenAI clarifies that GPT-4.5 is not considered a frontier AI model.
Today we’re releasing a research preview of GPT-4.5—our largest and best model for chat yet.
Rolling out now to all ChatGPT Pro users, followed by Plus and Team users next week, then Enterprise and Edu users the following week. pic.twitter.com/br5win5OEB
— OpenAI (@OpenAI) February 27, 2025
Expanded Access to GPT 4.5
Starting Thursday, GPT-4.5 will be accessible to subscribers of ChatGPT Pro, OpenAI’s $200-per-month plan, as part of a research preview. Developers subscribed to OpenAI’s paid API tiers will also gain access to the model. Plans are in place to extend availability to ChatGPT Plus and ChatGPT Team subscribers in the following week. An OpenAI spokesperson mentioned that broader access will be rolled out incrementally.
GPT-4.5 is a step forward in scaling up pre-training and post-training. By scaling unsupervised learning, GPT-4.5 improves its ability to recognize patterns, draw connections, and generate creative insights without reasoning. https://t.co/kFrvwwa2dk
— OpenAI (@OpenAI) February 27, 2025
GPT 4.5 was developed using the same foundational technique as previous GPT models, involving a significant increase in computing power and data during the unsupervised learning phase. This approach has historically led to substantial improvements in performance across various domains, including mathematics, writing, and coding. OpenAI notes that GPT 4.5 exhibits a deeper understanding of world knowledge and higher emotional intelligence. However, some benchmarks indicate that the benefits of scaling up data and computing resources may be reaching a plateau. For instance, GPT-4.5 underperforms compared to newer AI reasoning models from companies like DeepSeek and Anthropic on certain evaluations.
Operational Considerations
Operating GPT-4.5 is notably more expensive than its predecessors. OpenAI is currently assessing the long-term feasibility of offering GPT-4.5 through its API, considering the associated costs. The pricing structure for accessing GPT-4.5’s API is set at $75 per million input tokens (approximately 750,000 words) and $150 per million output tokens. In contrast, GPT-4o is priced at $2.50 per million input tokens and $10 per million output tokens. This significant cost difference reflects the increased computational demands of GPT-4.5.
Performance and Capabilities
While GPT 4.5 offers enhanced performance in certain areas, it is not intended to replace GPT-4o. The new model supports features such as file and image uploads and ChatGPT’s canvas tool but currently lacks support for ChatGPT’s realistic two-way voice mode. On OpenAI’s SimpleQA benchmark, which assesses models on straightforward factual questions, GPT-4.5 outperforms GPT-4o and OpenAI’s reasoning models, o1 and o3-mini, in terms of accuracy. Additionally, GPT-4.5 is reported to hallucinate less frequently, potentially reducing instances of generating incorrect information.
However, on coding-related benchmarks like SWE-Bench Verified, GPT-4.5’s performance is comparable to GPT-4o and o3-mini but falls short of OpenAI’s deep research models and Anthropic’s Claude 3.7 Sonnet. This suggests that while GPT-4.5 excels in certain tasks, it may not surpass specialized models in all areas.
Future Directions
The release of GPT-4.5 highlights the challenges associated with scaling AI models through traditional pre-training methods. OpenAI acknowledges that the anticipated gains from increasing data and computational resources are diminishing. In response, the company plans to integrate its GPT series with its ‘o’ reasoning series, starting with the upcoming GPT-5 model. This integration aims to enhance AI capabilities by combining the strengths of both model types. Despite the high training costs and operational challenges, GPT-4.5 serves as a crucial step toward developing more advanced AI systems.
Also see:
Indians Dominate H-1B Visas: 72% of Allocations Between Oct 2022 & Sept 2023
H-1B Visa Registration Opens on March 7; Fee Hike
Trudeau Calls Trump’s Canada Annexation Threat a “Real Concern”
Gautam Adani Donates Rs 10,000 Crore to Social Causes at Son’s Wedding
—————————————————————
It would mean the world to us if you follow us on Twitter, Instagram and Facebook