GPT-4.5 - The largest and best model for chat yet in GPT family
GPT-4.5 is our most advanced model yet, scaling up unsupervised learning for better pattern recognition, deeper knowledge, and fewer hallucinations. It feels more natural, understands intent better, and excels at writing, programming, and problem-solving.
Replies
reap
GPT-4.5 is definitely a step toward more intuitive, human-like AI—feels less like a tool and more like a thoughtful collaborator. It seems we moving into an era where AI adapts to users, not the other way around. Curious to see how different models carve out their unique strengths as subjective computing evolves.
It’s rolling out now for Pro users and API access. Priced in API at,
Input: $150/1M tokens ($37.50/1M for cached input)
Output: $150.00 / 1M tokens
Plus, Edu, and Enterprise users will get access next week.
How does it compare to other models released this week, such as Claude 3.7 and Grok-3? Is this a significant leap forward, or just the beginning of OpenAI’s push toward smarter, more context-aware models?
@usamaabidj don't know if it was the original plan, but it's really cool to see this one is not a step forward in the main benchmarks but something different. We need models that are better at driving conversations and adapting their tone, and this looks like the best ever at that!
fn7 Helix
@sama It's a tough day competing with you, but we're still holding strong at the top so far! @fn7 Helix
SocialBu
Amazing! How does GPT-4.5 handle bias in its unsupervised learning?
Depict
Nice! How does it compare to Claude 3.7?
@hambena According to what I read (from Every etc) GPT 4.5 is becoming like Claude 3.5 (more emotional intelligent) while Claude 3.7 is becoming more like GPT 4o - weird and as a fan of Claude, I don't like this. Anthropic should have just increased their input tokens (to Gemini levels) to have a bigger context window - that would have been the change I'd like to see and not becoming colder and more bot-like just to please coders. The vast majority of users aren't coders - maybe this is something that OpenAi realized...
@hambena @atmos_black I'm not 100% sure. but comparing token levels isnt 1 to 1. Depends on the engine connected to the context as well. If you compare o1 with 100-200k context with Gemini flash with 1-2m context. Sure Gemini looks good on paper. more context seems better. but the IQ you get from o1 with 10% of the context is way better. In my experience. That said. If you need to operate on vast amount of data, but dont need the extra IQ points. then Gemini might win out. Apples and Pears 😸
Can always rely on GPT
This release seems rather underwhelming to me. I mean they've improved logical processing but the price seems way too high to me 🤔 I feel like they're trying to establish themselves as the Apple of LLMs and just thinks people will pay no matter how it changes
hell nah bro this is too little compared to grok offering free deep research
All the best for the launch @usamaabidj & team!
Excited to give this a try and see how it stacks up against all the latest competition. Any idea when it will become available for Plus users?