Calvin Wankhede / Android Authority
TL;DR
- ChatGPT is now using GPT-5.5 Instant as its deafult model.
- Compared to GPT-5.3 Instant, OpenAI promises tighter replies and fewer hallucinations.
- ChatGPT is also learning to share where it’s pulling context from with memory sources.
With so many different AI options available to us, it’s understandable to feel a bit overwhelmed by choice. Beyond just picking between Gemini, OpenAI, Anthropic, and the rest, there’s usually also the question of which of their many models you want to use. Fast? Thinking? Pro? If all this is too much, and you tend to just go with the default, we’ve got some good news for you, as OpenAI updates ChatGPT so that its default model is now clearer and more correct.
Don’t want to miss the best from Android Authority?
Specifically, we’re talking about a model called GPT‑5.5 Instant. As you might remember, OpenAI first started announcing this family of GPT-5.5 models last month. And while those are all important, GPT‑5.5 Instant is arguably the most user-facing out of all of them, as it’s the one ChatGPT is going to employ by default — replacing the existing GPT‑5.3 Instant model.
Both of these are designed to be fast, but with GPT-5.5 Instant, OpenAI emphasizes just how much better it’s gotten at not being wildly incorrect — always a plus.
In one example, the company shares some handwritten math with ChatGPT, asking for feedback. While both models initially fail to spot the problem present in the equations, GPT-5.3 Instant ultimately throws its hands up in the air and declares the problem unsolvable, while GPT‑5.5 Instant keeps working and is eventually able to work out a solution.
Neither of those approaches is perfect, but that’s clearly an improvement. In attempting to quantify the gains, OpenAI says that GPT-5.5 Instant should produce 52.5% fewer hallucinations in response to “high-stakes” questions — stuff where your money or your life is on the line. In another assessment, inaccurate responses were down 37.3% compared to GPT-5.3 Instant.
Beyond just being less wrong, ChatGPT with GPT‑5.5 Instant also gets to the point a lot faster, not wasting time with long-winded answers, keeping things tight but also appropriately casual for this model — users can always tap GPT‑5.5 Pro or Thinking when they need greater insight, and detailed explanations.
You also may notice responses formed with GPT‑5.5 Instant doing a better job at remembering relevant context from past conversations — and delivering better quality output as a result. And with new support for memory sources, ChatGPT can show you exactly where all that context is coming from.
GPT‑5.5 Instant is available as of today as ChatGPT’s new default, and memory sources are beginning to roll out to supported plans.
Thank you for being part of our community. Read our Comment Policy before posting.