OpenAI said earlier versions, including GPT-5.2 Instant, would sometimes refuse questions that could have been answered safely or respond with what users described as overly cautious or preachy language.
With GPT-5.3 Instant, OpenAI said the model “significantly reduces unnecessary refusals” and tones down overly defensive or moralising preambles before answering.
OpenAI said the model is also available to developers through the API under the name “gpt-5.3-chat-latest”.
GPT-5.3 Instant reduces hallucination rates across multiple domains, including higher-stakes areas such as medicine, law and finance.
Ключевые утверждения в источнике B
GPT-5.3 Instant therefore represents an optimisation of that front-line model, the one users interact with most often.
The model sits within the broader GPT-5 architecture, where lighter “Instant” models handle the majority of traffic while deeper reasoning models are invoked for more complex requests.
The architecture behind “Instant” models OpenAI’s GPT-5 system is structured around a tiered model architecture.
Small adjustments to training data, alignment techniques, and response generation can therefore have an outsized effect on perceived quality.
Текстовые доказательства
Доказательства из источника A
ключевое утверждение
With GPT-5.3 Instant, OpenAI said the model “significantly reduces unnecessary refusals” and tones down overly defensive or moralising preambles before answering.
Ключевое утверждение, на котором строится подача материала.
ключевое утверждение
OpenAI said the model is also available to developers through the API under the name “gpt-5.3-chat-latest”.
Ключевое утверждение, на котором строится подача материала.
выборочный акцент
In internal evaluations focused on higher-stakes topics, OpenAI said GPT-5.3 Instant reduced hallucination rates by 26.8 per cent when using web access and by 19.7 per cent when relying onl…
Возможное выборочное акцентирование отдельных аспектов истории.
Доказательства из источника B
ключевое утверждение
GPT-5.3 Instant therefore represents an optimisation of that front-line model, the one users interact with most often.
Ключевое утверждение, на котором строится подача материала.
ключевое утверждение
The model sits within the broader GPT-5 architecture, where lighter “Instant” models handle the majority of traffic while deeper reasoning models are invoked for more complex requests.
Ключевое утверждение, на котором строится подача материала.
Доказательства паттернов искажения
Источник A · Эффект фрейминга
In internal evaluations focused on higher-stakes topics, OpenAI said GPT-5.3 Instant reduced hallucination rates by 26.8 per cent when using web access and by 19.7 per cent when relying onl…