OpenAI’s GPT -5.3 Instant Rolls Out, Promises Accuracy, Fewer Rejections

The core model used throughout ChatGPT has been updated by the latest GPT-5.3 Instant version. OpenAI has shared the improvements the new model will provide, and they include:

1. Response relevancy
2. Better conversational flow
3. Improved tone in daily interactions
4. Strengthened accuracy when it comes to factual and web-based answers

Users will also be pleased to learn that refusals will be reduced and so will overly cautious responses for sensitive and generalised user queries.

Also read: WhatsApp Releases Update Allowing Group Message History (Across Devices for New Members) 

Enhanced Judgement Around Sensitive Queries

OpenAI has stated a major upgrade in the GPT -5.3 Instant is its ability to cut out dead ends by providing helpful query-focused answers.

The company says ChatGPT powered by the GPT -5.3 Instant update will work to offer an appropriate answer that’s not overly cautious and addresses the query without being overly defensive or moralizing.

Boosting Factuality

This update will deliver more factual answers with reduced hallucinations for queries across all topics. The company chose to measure factuality on two internal benchmarks:

“one focused on higher-stakes domains such as medicine, law, and finance, and another measuring hallucination rates on de-identified ChatGPT conversations that users flagged as factual errors”

The factuality increased by:

• 26.8% reduction in hallucinations with web use and 19.7% reduction in hallucinations without web access for higher-stakes domains (finance, health, and law).

• 22.5% hallucination reduction with web use and 9.6% reduction when there was no web access for an evaluation based on user-flagged factual errors.

Well-Synthesized and Helpful Responses

Queries requiring internet research will yield better answer quality courtesy of the latest update.

The responses are a blend of research results, logical reasoning, and own knowledge to deliver a result that’s been contextualized. There will also be less of the long lists of content and sources that aren’t closely connected to the query or suitable answer.

Stronger Storytelling Capabilities

When it comes to testing the writing abilities of ChatGPT, the -5.3 Instant model will improve these capabilities, especially when writing fiction, refining content, and improving expressive pose.

The model seamlessly transitions between creative and practical queries and maintains structuring and clarity throughout the process.

Limitations of GPT -5.3 Instant

Despite the GPT -5.3 Instant, OpenAI admit there is work to be done as some limitations exist in this model:

1. Non-English language responses (e.g Japanese and Korean) seem literal and work needs to be done for optimal tone and naturalness.
2. Feedback will be monitored to continue tone improvements and customization options.

GPT -5.3 Instant Availability

The new ChatGPT -5.3 Instant update will immediately be available to all ChatGPT users. The older -5.2 Instant model will be available to paid users for three months and can be accessed under the Legacy Models section to select models. 

This older model will be permanently retired on 3 June 2026.
The update is available to developers in the API as ‘gpt-5.3-chat-latest.’ Note that Updates are still incoming for Thinking and Pro models.