GPT-5.3 Instant cuts hallucinations by 26.8% as OpenAI shifts focus from speed to accuracy
Source: VentureBeat
OpenAI GPT‑5.3 Instant Overview
- Reduced hallucinations – Up to 26.8 % fewer hallucinations compared with the previous model, emphasizing accuracy and conversational reliability over raw performance gains.
- Enhanced user experience – Improves tone, relevance, and overall conversation flow while generating fewer refusals.
- Availability – The Instant model is the default for ChatGPT users and is also accessible via the API.
- Future updates – Currently, only the Instant model has been upgraded to 5.3. OpenAI plans to roll out the 5.3 update to the other models in the ChatGPT, Thinking, and Pro families “soon.”
GPT‑5.3 Instant Cuts Hallucinations by Up to 26.8 %
OpenAI ran two internal evaluations:
- Higher‑stakes domains – medicine, finance, and law.
- User‑feedback study – based on real‑world queries.
Key Findings
| Evaluation | Metric | Result |
|---|---|---|
| Higher‑stakes (web) | Hallucination reduction | ‑26.8 % |
| Higher‑stakes (internal knowledge) | Reliability increase | +19.7 % |
| User‑feedback (web search) | Hallucination reduction | ‑22.5 % |
OpenAI says GPT‑5.3 Instant is more reliable because it balances information from the internet with its own internal training and reasoning more effectively.
“More broadly, GPT‑5.3 Instant is less likely to over‑index on web results, which previously could lead to long lists of links or loosely connected information. It does a stronger job of recognizing the subtext of questions and surfacing the most important information, especially upfront, resulting in answers that are more relevant and immediately usable, without sacrificing speed or tone.” – OpenAI
Example
When a user asks about the biggest signing in Major League Baseball and its impact, the previous model (GPT‑5.2) often defaulted to summarizing search results. GPT‑5.3 Instant instead provides a concise, context‑aware answer.
Accuracy Overtakes Performance as OpenAI’s Selling Point
With this release—first on its most‑used model—OpenAI wants enterprise customers and ChatGPT users to see that the battlefront is no longer just about raw performance (speed, token efficiency) but about adherence to factual information.
Industry Context
- Anthropic – announced that its new Claude Sonnet 4.6 has fewer hallucinations.
- Google – was forced to pull its Gemma 3 model after it hallucinated false information about a lawmaker.
These moves underscore a broader shift toward reliability and factual correctness as the primary differentiators for large‑language‑model providers.
GPT‑5.3 Instant Dials Back Refusals and “Cringe” Tone
“This update focuses on the parts of the ChatGPT experience people feel every day: tone, relevance, and conversational flow. These are nuanced problems that don’t always show up in benchmarks, but shape whether ChatGPT feels helpful or frustrating. GPT‑5.3 Instant directly reflects user feedback in these areas,” — OpenAI blog post
What’s new?
- More natural conversation style – The model moves away from the “cringe” tone that many users described as over‑bearing and overly assumptive about intent.
- Consistent personality – OpenAI says the chat platform’s personality will stay stable across updates, preventing sudden tonal shifts.
- Fewer refusals – The previous version often declined to answer questions that didn’t violate guardrails, or responded with overly cautious, preachy language. GPT‑5.3 Instant is designed to answer directly, without unnecessary caveats.
Key improvements
| Area | Previous behavior | GPT‑5.3 Instant change |
|---|---|---|
| Tone | “Cringe” / overbearing, moralizing preambles | Natural, conversational, consistent |
| Refusals | Frequent, even for benign queries | Significantly reduced; answers provided unless a guardrail is truly triggered |
| Caveats | Overly defensive or preachy explanations | Direct answers with minimal preamble |
Remaining limitations
- Language coverage – Answers in Korean and Japanese can still sound stilted or unnatural.
- Edge cases – While refusals are reduced, the model will still decline requests that truly violate policy.
Sources
- OpenAI announcement:
- Coverage of the “cringe” tone issue:
Safety Card Shows Regressions in Sexual Content and Self‑Harm Categories
OpenAI’s spokesperson told VentureBeat that the new model does not support adult content, as the company is still figuring out “how to maximize user freedom while maintaining our high safety bar.” No timeline has been provided for when this functionality might be released.
OpenAI performed safety benchmarking on the new model and published the results on its safety card. The key findings are:
- Overall performance: The model performed well against disallowed content but did not reach the safety level of GPT‑5.2 Instant. OpenAI noted that these results could change after launch.
- Regressions:
“GPT‑5.3 Instant shows regressions relative to GPT‑5.2 Instant and GPT‑5.1 Instant for disallowed sexual content, and relative to GPT‑5.2 Instant for self‑harm on both standard and dynamic evaluations.”
- Other categories: In most other safety categories, the model performs on par with or better than previous releases. The observed regressions for graphic violence and violent illicit behavior have low statistical significance.
Expect a New Model Soon?
After announcing GPT‑5.3 Instant and noting that updates for Thinking and Pro will be coming soon, OpenAI teased that even this new model could be retired.
In a post on X, OpenAI said GPT‑5.4 is coming “sooner than you think.”
OpenAI did not elaborate on:
- What changes, if any, we can expect with GPT‑5.4
- Which modes will receive it first
GPT‑5.2 Instant, the predecessor model, will remain available in the ChatGPT model picker until June 3, when it will be retired.