OpenAI announced the release of GPT‑5, a new language model that delivers 30% faster inference speeds and a 25% reduction in hallucinations compared to GPT‑4, according to a blog post from the company.
OpenAI has announced the launch of GPT‑5, its latest generative‑AI model, promising notable improvements in speed and reliability over its predecessor.
Key performance upgrades
According to OpenAI’s blog post, GPT‑5 delivers roughly 30% faster inference than GPT‑4, reducing latency for real‑time applications.
The company also claims a 25% reduction in hallucinations, meaning the model generates fewer factually inaccurate statements.
Technical enhancements
OpenAI attributes the speed gains to a redesigned transformer architecture and optimized kernel operations that better leverage modern GPU hardware.
To curb hallucinations, the team introduced a more rigorous fine‑tuning pipeline that incorporates additional factuality checks and a larger curated dataset of verified information.
Potential impact and use cases
Developers can expect quicker response times in chatbots, code assistants, and other interactive tools, while the lowered hallucination rate aims to improve trust in high‑stakes domains such as medical advice and legal drafting.
- Enhanced virtual assistants
- More reliable content generation
- Improved real‑time translation
- Better support for enterprise workflows
We’re focused on making AI both faster and more trustworthy for everyone.
Read the full announcement on OpenAI’s blog: Read the report.