Imminently, a ChatGPT ‘router’ that automatically selects OpenAI models for your job

Subscribe to our weekly newsletter to receive only the most important information for enterprise AI, data and security leaders. Submit Now

Is there help on the horizon? OpenAI appears to be delivering help soon. Reports from AI influencers and OpenAI’s researcher on X have been appearing in the past few days. There @tszzl)” (speculated as being technical team member Tarun Gagineni), of a new “router”, function that will automatically choose the best OpenAI models to respond to a user’s input, depending on its content.

Roon responded to Lisan al Gaib, an influencer, who said that she “didn’t want a router, I wanted to be able select the models I used” on the social network X, yesterday, July 20,2025:

You’ll still have the option to select. This product is to ensure that doctors don’t get stuck on 4o mini”

Similarly Yuchen Jin, Cofounder & CTO at AI inference cloud provider Hyperbolic Labs. On July 19, X posted a message.

I heard that GPT-5 will be imminent from a little bird.

• There are multiple models, not just one. It has a router which switches between reasoning, tool-using, and non-reasoning models.
• Sam said that they would “fix model naming”. Prompts will simply auto-route to model.
• GPT-6 has been trained.

Just hope they don’t delay it for safety tests. 🙂“

A presumably more advanced GPT-5 would be a huge news item if it were released. However, the router could make life easier and more intelligent to the average ChatGPT user.

The router would also follow other third-party products, such as the web based Token Monster Chatbot, which automatically selects and combines responses from multiple third party LLMs to answer user queries. OpenAI’s spokesperson declined to comment on the router idea or “Roon”‘s comments.

Solving the overabundance problem

OpenAI has always shared in a blog or release note, or both, what it believes that particular model can help with.

OpenAI’s reasoning models from the “o” series — o3, O4-mini, and O4-mini High — have performed better in math, science, and code tests, thanks to benchmarking tests. Non-reasoning models, like the new GPT 4.5 and 4.1, seem to perform better at creative writing and communication tasks.

Power users and AI influencers may know very well what these models are good at and not so great at. Regular users, who may not be as familiar with the industry, nor have the resources to test all of them on the same input prompts to compare the results, will struggle to make sense out of the confusing array of options.

This could mean that they are missing out on ChatGPT’s more intelligent or capable responses for their task. In fields such as medicine, Roon said, the difference can be life or death.

I find it interesting to speculate about how an automatic LLM-router might change public perceptions and adoption of AI in general. ChatGPT had 500 million active users by March. If more people were guided to more intelligent and capable LLMs for their AI queries, then the impact of AI would be felt more acutely on their workloads, and the global economy as a whole, creating a “snowball effect”.

As more people see the benefits of ChatGPT automatically selecting the right AI model to answer their queries and as more businesses reap greater efficiency, they will be more convinced of the utility and willing to pay for AI.

Right now, however, this is being held back by the fact that ChatGPT’s model picker requires users to A. be aware of the models available and B. be informed about what they are good for. The process is still manual. ChatGPT users are currently overwhelmed by choice.

I hope that any hypothetical OpenAI router will seamlessly guide them to the right product for their needs when they need it, like a trusted shopkeeper showing up and freeing you from product paralysis.