Source URL: https://simonwillison.net/2025/Feb/27/introducing-gpt-45/#atom-everything
Source: Simon Willison’s Weblog
Title: Introducing GPT-4.5
Feedly Summary: Introducing GPT-4.5
GPT-4.5 is out today as a “research preview" – it’s available to OpenAI Pro ($200/month) customers but also to developers with an API key.
OpenAI also published a GPT-4.5 system card.
I’ve started work adding it to LLM but I don’t have a full release out yet. For the moment you can try it out via uv like this:
uvx –with ‘https://github.com/simonw/llm/archive/801b08bf40788c09aed6175252876310312fe667.zip’ \
llm -m gpt-4.5-preview ‘impress me’
It’s very expensive right now: currently $75.00 per million input tokens and $150/million for output! For comparison, o1 is $15/$60 and GPT-4o is just $2.50/$10.
As far as I can tell almost all of its key characteristics are the same as GPT-4o: it has the same 128,000 context length, handle the same inputs (text and image) and even has the same training cut-off date of October 2023.
So what’s it better at? According to OpenAI’s blog post:
Combining deep understanding of the world with improved collaboration results in a model that integrates ideas naturally in warm and intuitive conversations that are more attuned to human collaboration. GPT‑4.5 has a better understanding of what humans mean and interprets subtle cues or implicit expectations with greater nuance and “EQ”. GPT‑4.5 also shows stronger aesthetic intuition and creativity. It excels at helping with writing and design.
They also report a SimpleQA hallucination rate of 37.1% – a big improvement on GPT-4o (61.8%) and o3-mini (80.3%) but not much better than o1 (44%). The coding benchmarks all appear to score similar to o3-mini.
OpenAI don’t seem to have enormous confidence in the model themselves:
GPT‑4.5 is a very large and compute-intensive model, making it more expensive than and not a replacement for GPT‑4o. Because of this, we’re evaluating whether to continue serving it in the API long-term as we balance supporting current capabilities with building future models.
It drew me this for "Generate an SVG of a pelican riding a bicycle":
Accessed via the API the model feels weirdly slow – here’s an animation showing how that pelican was rendered – the full response took 112 seconds!
Tags: llm, generative-ai, uv, openai, ai, llms, pelican-riding-a-bicycle
AI Summary and Description: Yes
Summary: The introduction of GPT-4.5 marks a significant development in the landscape of generative AI models, offering improved features such as enhanced human collaboration, creativity, and a reduced hallucination rate. Its pricing model, while steep, reflects its computational demands, indicating a sustained evolution in large language models (LLMs) relevant to AI security and infrastructure.
Detailed Description: The text discusses the launch of GPT-4.5, a new release from OpenAI that builds upon the capabilities of its predecessor, GPT-4o. Here are the major points:
– **Availability**: GPT-4.5 is available as a research preview for OpenAI Pro users and developers possessing an API key, with a notable subscription fee of $200 per month.
– **System Card**: OpenAI has published a system card detailing the characteristics and functionalities of GPT-4.5.
– **Cost**: The operation cost for GPT-4.5 is considerably high—$75 for every million input tokens and $150 for output tokens, in contrast to the significantly lower costs of its predecessors (e.g., o1, GPT-4o).
– **Performance Characteristics**:
– Maintains a context length of 128,000, capable of handling both text and image inputs.
– The training cut-off date remains consistent with GPT-4o (October 2023).
– **Improvements Over Previous Models**:
– The model exhibits enhanced understanding and emotional quotient (EQ), allowing it to interpret human cues and implicit expectations better.
– It showcases a notable improvement in creativity and aesthetic judgment, particularly in writing and design tasks.
– The hallucination rate stands at 37.1%, marking a significant improvement over GPT-4o and other preceding models.
– **Performance and Confidence**:
– Despite the advancements, OpenAI expresses caution regarding the model’s practicality and sustainability, indicating that while it is powerful, it is not designed as a replacement for GPT-4o due to its high computational demands.
– There is uncertainty about future API support for GPT-4.5 as OpenAI weighs sustaining current models against developing new ones.
– **User Experience**: The slower response times when accessed through the API may hinder user experience.
This launch introduces critical implications for professionals in AI and security fields, particularly around the cost and effectiveness of increasingly complex models. The enhancements in understanding and creativity also provoke discussions about the ethical use and reliability of such generative AI systems in production environments.