ChatGPT’s default model is now more direct, more clear, and less wrong about things

Source: Android Authority

Calvin Wankhede / Android Authority

TL;DR

• ChatGPT now uses GPT‑5.5 Instant as its default model.
• Compared to GPT‑5.3 Instant, OpenAI promises tighter replies and fewer hallucinations.
• ChatGPT can now show the memory sources it uses to pull context.

With so many AI options—Gemini, OpenAI, Anthropic, and others—it’s easy to feel overwhelmed. Beyond choosing a provider, you often have to decide which of their many models to use: Fast? Thinking? Pro? If you usually stick with the default, there’s good news: OpenAI has updated ChatGPT so its default model is now clearer and more correct.

Introducing GPT‑5.5 Instant

OpenAI announced the GPT‑5.5 family last month. Among them, GPT‑5.5 Instant is the most user‑facing, replacing the previous GPT‑5.3 Instant as ChatGPT’s default model.

Both models are designed for speed, but GPT‑5.5 Instant is marketed as significantly less prone to major errors.

Improved Accuracy

In a test where handwritten math was submitted for feedback, both models initially missed the problem. GPT‑5.3 Instant eventually gave up, labeling the problem unsolvable, while GPT‑5.5 Instant kept working and eventually solved it.

OpenAI quantifies the gains:

• 52.5% fewer hallucinations on “high‑stakes” questions (where money or life is on the line).
• 37.3% fewer inaccurate responses compared to GPT‑5.3 Instant.

Faster, More Concise Replies

GPT‑5.5 Instant delivers tighter, more to‑the‑point answers without unnecessary verbosity, while still maintaining a casual tone appropriate for the model. Users can still switch to GPT‑5.5 Pro or Thinking modes when deeper insight or detailed explanations are needed.

Better Context Handling

Responses from GPT‑5.5 Instant show improved memory of relevant context from past conversations, resulting in higher‑quality output. New support for memory sources lets ChatGPT display exactly where the context originates.

Availability

GPT‑5.5 Instant is live as the default model for ChatGPT, and memory sources are rolling out to supported plans.