Want smart insight into your inbox? Enterprise AI, only what matters to data and security leaders, sign up for our weekly newspapers. Subscribe now
Openai has increased continuously in 2.5 years since the onset of Chatgpt, the number of big language models (LLMS) that the company has provided as an alternative to powering its hit chatbot.
In fact, now a total of 7 (!!!) are different AI models that pay chat customers ($ 20 plus tier and more expensive levels), when you can choose when interacting with reliable chatbott – each with its strength and weaknesses.
This includes:
- GPT-4o
- O3
- O4-Mini
- O4-Mini-High
- GPT -4.5 (Research Preview)
- GPT-4.1
- GPT -4.1 -Mune
But how a user should decide Which one To use their special signs, questions or tasks? After all, you can only choose one at once.
AI Impact series returns to San Francisco – 5 August
The next phase of AI is here – are you ready? Leaders of Block, GSK and SAP include how autonomous agents are re-shaping the enterprise workflows-from the decision making of time-to-end and automation.
Now secure your location – space is limited:
Help on the way?
Help on the way to the rich way from OpeniI seems to be on the X from AI’s influencers in the last few days, including OpenaiI’s own researchers “Raon ,On @tszzl x) “(Technical team member Tarun Goginani is estimated) – for a new” router “function that will automatically choose the best Openai model to respond to the user’s input on the fly, based on the content of the specific input.
As Roon posted on social network X on yesterday, July 20, 2025, after the response to impressing Lisan al -Gab’s statement that they do not want a model router that I want to be able to select the model I want to be able to choose what I use “:
“You will still be able to select. It is a product to ensure that doctors are not stuck on 4O-mini”
Similarly, the CTO of Hyperbolic Labs, co-founder and AI Invention Cloud Provider Hyperbolic Labs, Wrote in an X post on 19 July.
,Hurd GPT -5 is adjacent, from a small bird.
- It is not a model, but there are many models. It has a router that switchs between logic, non-generation and tool-oining models.
- So Sam said that they “will fix the model naming”: The signal will just auto-root for the right model.
- GPT-6 is in training.
I hope they are not delaying it for more security tests. ,,
Whereas probably more advanced GPT-5 models (and will be) will be huge news. If and when released, the router average Chatgpt can make life much easier and more intelligent for the customer.
This web-based token will also follow on heels of other third-party products such as Monster Chatboats, which automatically choose and combine reactions from many third-party LLM to answer users questions.
Asked about the idea of the router and the comments from “Roon”, an Openai spokesman refused to provide an response or more information at this time.
Solid the problem of choice
To be clear, every time Openai has released a new LLM for the public, it has either shared or released notes in a blog post with hard work or both think it is good for special models and designed to help with it.
For example, Openai’s “O” series Reasoning Models-O3, O4-Min, O4-Mini High- have performed better on mathematics, science and coding tests for benchmarking tests, while non-purpose models such as the new GPT-4.5 and 4.1 have improved in creative writing and communication functions.
Dedicated AI -affected and electric users can understand very well that all these different models are good and not so good.
But regular users who do not follow the industry closely, nor are the time and finance available are available to all those available to test on the same input signals and compare the output, will struggle significantly to create an understanding of the horrific array of options.
This may mean that they are remembering more smarter, more intelligent, or more competent reactions than chat for their work at hand. And in the case of fields such as drugs, as Roon said, the difference may be one of life or death.
It is also interesting how an automatic LLM router can change public assumptions and adopt AI more widely.
Chatgpt already counted 500 million active users as March. If more of these people were automatically directed to the more intelligent and competent LLM to handle their AI questions, then the effect of AI on their workload and the effect of the entire global economy would be likely to be felt more intense, leading to a positive “snoball” effect.
That’s, because more people saw more benefits than chat Automatically Choosing the right AI model for their questions, and as more enterprises gained more efficiency from this process, more and more individuals and organizations would be convinced by the usefulness of AI and would be more inclined to pay for it, and even as they did, even, More The AI-operated workflow will spread to the world.
But right now, it is probably being held a little back from the fact that the Chatgpt model picker requires a user. It is also known that they also have a model option and B has some levels of informed awareness about what is good for these models. All this is still a manually operated process.
Like going to the supermarket in your city and staring at the corridors of grains and different sauce, the average Chatgpt user currently encounters with an excess of choice.
It is expected that any fictional openi router basically helps them direct them to the right model product when they need it – such as showing a trusted shopkeeper to free his product paralysis.
