OpenAI has just introduced GPT-4.5, its latest large language model. CEO Sam Altman describes it as the first AI that feels like talking to a “thoughtful person.” However, this model marks a significant shift—OpenAI confirms that GPT-4.5 will be the last non-reasoning model. The next step? Merging the GPT-series and o-series into a unified AI system, likely GPT-5.
But what does GPT-4.5 bring to the table? How does it compare to previous models, and where exactly does this fit in the OpenAI’s arsenal? Let’s break it all down.
What is GPT-4.5?
GPT-4.5, internally codenamed ‘Orion,’ is OpenAI’s latest large language model, an upgrade over their previous GPT models like GPT 4o. This is designed to improve conversation quality, writing assistance, and problem-solving. It’s bigger and more expensive to run than any previous OpenAI model and has been trained with more data and computing power than ever before.
ChatGPT 4.5 model can be summarized down to these key highlights:
- More Human-Like Conversations: OpenAI says this model has a “refined personality” and responds in a more natural, intuitive way.
- Better Writing and Programming Support: GPT-4.5 excels at improving writing, debugging code, and automating tasks.
- Stronger World Knowledge: It understands information more deeply, making it better at answering factual questions.
- Fewer Hallucinations: OpenAI claims GPT-4.5 is more reliable and less likely to generate false information than previous models.
Despite being the latest model, it definitely has many limitations or weaknesses:
- Not the Best for Logical Reasoning: This is not an O-series model and does not have extended reasoning capabilities. This model is designed for quick answers and underperforms compared to models like DeepSeek’s R1, Anthropic’s Claude 3.7 Sonnet, and OpenAI’s own o3-mini.
- Extremely Expensive: GPT-4.5 costs $75 per million input tokens and $150 per million output tokens, compared to just $2.50 and $10 for GPT-4o. OpenAI is actually considering whether to continue offering it to devs.
- Limited Features in ChatGPT: It doesn’t support multimodal features like voice mode, video, or screen sharing (yet).
- Delayed Rollout: GPU shortages mean it’s only available to Pro users now, with Plus users getting access later.
Also Read:
- OpenAI Rolls Out Deep Research to All Paid Users – But Is It Enough to Beat Gemini
- Elon’s xAI Grok 3 Is Here: How Does It Compare to ChatGPT-4o and Gemini 2.0?
How Does GPT-4.5 Perform in Benchmarks?
Benchmarks help us measure how well AI models perform in different areas. Here’s how GPT-4.5 stacks up against both OpenAI’s previous models and its competitors:
GPT-4.5 shows solid improvements over GPT-4o in factual accuracy, scoring 62.5% on the SimpleQA benchmark compared to GPT-4o’s 38.2% and OpenAI’s o1 at 47%. It also has a lower hallucination rate (37.1%) than GPT-4o (61.8%) and o1 (44%), though models like DeepSeek’s R1 and Perplexity’s Deep Research still do better at fact-checking. When it comes to coding, GPT-4.5 performed better than GPT-4o and o3-mini on OpenAI’s SWE-Lancer benchmark but lagged behind Anthropic’s Claude 3.7 Sonnet and OpenAI’s deep research models.
However, it struggles with complex problem-solving, scoring just 36.7% on AIME (math) compared to o3-mini’s 87.3%, and 71.4% on GPQA (science), trailing DeepSeek’s R1. On the bright side, it beats GPT-4o in multilingual (85.1% vs. 81.5%) and multimodal (74.4% vs. 69.1%) tasks, but it still lacks full multimodal capabilities like voice and video support.
Human Preference Ratings is a method used by OpenAI to evaluate how real users perceive the quality of responses from different AI models. In this rating, Users preferred GPT-4.5 over GPT-4o for creative writing, professional queries, and everyday conversations, but it still trails behind models like Claude 3.7 Sonnet in structured reasoning and legal document drafting.
Overall, GPT-4.5 is a step up in factual knowledge and conversational abilities, but when it comes to deep reasoning and structured problem-solving, models like DeepSeek’s R1 and Claude 3.7 Sonnet still have the edge.
How to Access GPT-4.5
Starting today, ChatGPT Pro users ($200/month) can access GPT-4.5. OpenAI plans to roll it out to ChatGPT Plus ($20/month) and Team ($30/month) users next week once more GPUs are available. Users can try GPT 4.5 by just picking that model from the model picker.
For developers, GPT-4.5 is being made available through OpenAI’s API, including the Chat Completions API, Assistants API, and Batch API. It supports key features like function calling, structured outputs, streaming, system messages, and image inputs, making it a versatile tool for various AI-driven applications. However, it currently does not support multimodal capabilities such as voice mode, video, or screen sharing.
GPT-4.5 is an exciting step forward, especially for writing, general conversation, and factual accuracy. However, it falls short in deep reasoning and structured problem-solving, which the company strongly hints will not be the case from later models.
