GPT-6.1 Sol: OpenAI's New AI at a Fraction of the Cost
By Chatday Editorial Team ·
OpenAI has a new model, and the interesting part is not that it is smarter. It is that it is a lot cheaper.
GPT-6.1 Sol arrived on September 29 at OpenAI’s DevDay. The headline, in plain words: it is nearly as capable as OpenAI’s most powerful model, but it runs at roughly a fifth of the price. And you do not need to wait for it to show up in the ChatGPT app. You can already chat with it today in Chatday, right next to Claude, Gemini and Grok.
What GPT-6.1 Sol actually is
OpenAI does not sell one single model. It offers a range of them. At the top sits GPT-6 Astra, the most powerful and most expensive option. GPT-6.1 Sol is the new model that sits just below it: an everyday model meant to do most of the same jobs for far less money.
The name sounds technical, so here is the simple version. “GPT-6.1” is the generation, a small step up from the earlier GPT-6 Sol. “Sol” is just the label OpenAI gives this tier of model. You do not need to remember any of this. What matters is what it can do and what it costs.
OpenAI describes Sol as good at practical, hands-on work: writing and debugging code, working its way through software and apps step by step, and handling longer professional tasks without losing the thread. If you have ever wished an assistant could not just answer a question but actually carry out a multi-step job, that is the direction this model is pointed.
Why “a fifth of the price” is the real story
New models usually sell themselves on being smarter. Sol is doing something different, and it is arguably more useful for normal people: it is making near-top AI cheap enough to use freely.
OpenAI says Sol nearly matches Astra, its most powerful model, on the kinds of tasks most people care about, while costing about one-fifth as much to run. For developers paying per use, that is a large bill cut. For everyone else, the effect is simpler. When a capable model is cheap to run, you get to use it more often without worrying about the cost every time.
This is part of a wider pattern worth watching. The headline models get the attention, but the quieter releases, the cheaper ones that do most of the job, are what most people end up using. We compared that trade-off in our look at fast and smart AI models, and Sol is a clear example of it.
What it is good at, and where it is not for you
Sol is aimed squarely at work: coding, repetitive digital tasks, and longer professional jobs. If that is what you need, it is a strong, affordable pick.
But be honest about the fit. If you mostly want help writing an email, planning a trip, summarizing a long document or brainstorming ideas, you do not need a coding-focused model. A general chat model handles those just as well, and often feels friendlier for casual back-and-forth. Picking the right tool for the task beats always reaching for the newest name.
There is also a catch on access. Right now Sol is showing up inside OpenAI’s own work and developer products first. It is not in the standard ChatGPT chat box yet, so if you went looking for it there, you might not find it. That is exactly the kind of gap an all-in-one platform closes: in Chatday, the model is simply there to pick.
Which new model should you actually use?
The honest answer is that there is no single winner anymore, and that is good news. Different models are better at different things, and the smart move is to match the model to the moment. Here is a simple guide to models you can open today:
| Model | Best for | Good to know |
|---|---|---|
| GPT-6.1 Sol | Coding and longer hands-on tasks | New, capable, low cost |
| Claude Opus 5 | Careful writing and reasoning | Thoughtful and thorough |
| Gemini 3.1 Pro | Research and mixed text-and-image jobs | Fast, broad, well-rounded |
| Grok 4.6 | Casual, up-to-the-minute questions | Relaxed and conversational |
The point is not to crown one of them. It is that you should not be locked into a single app’s single model. Want to see two of them answer the same question side by side before you commit? Compare them in the model comparator.
If you have been weighing whether a single ChatGPT subscription is worth it, a launch like this is a good reminder that the best model for you this week might come from a different lab entirely.
The fastest way to see whether Sol fits your work is to give it a real task and watch how it handles it. Open it, paste in something you are actually stuck on, and if it is not the right fit, switch models in the same window.