GPT-5.5 Instant Cuts Hallucinations by 52.5% and Delivers More Concise Answers

OpenAI's free GPT-5.5 Instant replaces GPT-5.3 as the default model, slashing hallucinations by 52.5% in high‑risk domains, improving factual accuracy, providing shorter yet precise responses, adding memory‑controlled personalization, and rolling out to all ChatGPT users via the chat‑latest API.

SuanNi
SuanNi
SuanNi
GPT-5.5 Instant Cuts Hallucinations by 52.5% and Delivers More Concise Answers

OpenAI officially launched GPT-5.5 Instant, making it the default model for all ChatGPT users and offering it for free. The new model supersedes GPT-5.3 Instant and is exposed in the API as chat-latest, with the older version retained for three months before retirement.

Internal evaluations show that in high‑risk prompts covering medicine, law, and finance, GPT-5.5 Instant generates hallucinated statements 52.5% fewer times than GPT-5.3 Instant. In user‑flagged challenging dialogues, the rate of inaccurate statements drops by 37.3%.

The model also delivers more accurate and concise answers. In a demonstrated algebra problem, GPT-5.5 initially proposes an incorrect solution, then substitutes the answer back into the original equation, detects the substitution failure, identifies the user's algebraic mistake, and recomputes the correct result using the quadratic formula. By contrast, GPT-5.3 detects the failure but stops early, incorrectly concluding that no real solution exists and never revisits the calculation.

Compared with the older version, GPT-5.5 reduces the word count of its responses by 26.6% and the line count by 36.4%, eliminating redundant phrasing and unnecessary emojis while preserving a friendly tone.

Memory control is a core upgrade: with user consent, the model can retrieve past conversations, local files, and Gmail context. The new “Memory sources” panel displays which memories influence a response, allowing users to delete or edit irrelevant items. When sharing a conversation, memory sources are automatically hidden, and a temporary chat mode can disable memory usage entirely.

Personalization is further enhanced. For example, when giving travel or dining recommendations, GPT-5.5 leverages stored preferences and prior chats to provide highly tailored suggestions, whereas GPT-5.3 offers generic, broad recommendations based only on the user's current location.

The rollout is phased: GPT-5.5 Instant is being pushed to all ChatGPT users, with enhanced memory and personalization features first available to Plus and Pro web users, followed by mobile, free, Go, Business, and Enterprise tiers in the coming weeks.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

personalizationAIOpenAIhallucination reductionGPT-5.5memory control
SuanNi
Written by

SuanNi

A community for AI developers that aggregates large-model development services, models, and compute power.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.