OpenAI has updated its large language model (LLM) in ChatGPT to GPT-5, which it says takes a significant step towards artificial general intelligence (AGI). In a blog post, the company said GPT-5 delivers leaps in accuracy, speed, reasoning, context recognition, structured thinking and problem-solving.
“We anticipate early adoption to drive industry leadership on what’s possible with AI powered by GPT‑5, leading to better decision-making, improved collaboration and faster outcomes on high-stakes work for organisations,” said OpenAI.
From a technology perspective, OpenAI has built GPT-5 around a unified system, which it claims offers a smart, efficient model that answers most questions, combined with a deeper reasoning model for harder problems and a real‑time router that can quickly decide which to use. The company said the router is continuously trained on real signals, including when users switch models, preference rates for responses and measured correctness.
One of the features of the router, according to OpenAI, is that it enables the LLM to continue operating using a smaller model, once usage limits are reached.
Another change concerns safety. With GPT‑5, OpenAI has introduced a new form of safety training called “safe completions”, which teaches the model to give the most helpful answer where possible, while still staying within safety boundaries.
Describing the approach, the company said that sometimes GPT-5 may offer a partial answer to a user’s question or only answer at a high level. “If the model needs to refuse, GPT‑5 is trained to transparently tell you why it is refusing, as well as provide safe alternatives,” said OpenAI.
It also claims to offer better AI coding compared to previous models. “We’ve found GPT‑5 is excellent at digging deep into codebases to answer questions about how various pieces work or interoperate,” said OpenAI. “In a codebase as complicated as OpenAI’s reinforcement learning stack, we’re finding that GPT‑5 can help us reason about and answer questions about our code, accelerating our own day-to-day work.”
In the SWE-Bench Verified, a model is given a code repository and issue description, and must generate a patch to solve the issue. GPT-5 achieved an accuracy of 75% with around 10,000 tokens, compared with OpenAI’s o3 model, which scored 69% accuracy using 13,741 tokens.
We’re at a tipping point. GPT-5 promises even more realism, more precision and more ease for the user. That’s great for innovation, but it’s also a gift to fraudsters Gary Hall, Medius
Commenting on the launch, Grant Farhall, chief product officer at Getty Images, said GPT-5 would further reshape its relationship with content, creativity and imagery.
“As AI content becomes more convincing, we need to ask ourselves, ‘Are we protecting the people and creativity behind what we see every day?’ Authenticity matters, but it doesn’t come for free. It’s more important now that we look at exactly how AI models are being trained – if it is on permissioned content and that creators are being compensated for their works being trained,” he said.
“In addition, our global consumer research has found that people increasingly crave authenticity and transparency, especially in visual content. With AI evolving at pace, the real question is: Will GPT-5-generated content feel relatable and real, or will it further fuel demand for genuinely human, nuanced work?”
As AI gets closer to AGI, the risk of such systems being used fraudulently also increases significantly.
Gary Hall, chief product Officer at Medius, warned: “We’re at a tipping point. GPT-5 promises even more realism, more precision and more ease for the user. That’s great for innovation, but it’s also a gift to fraudsters. When AI-generated documents are indistinguishable from the real thing, legacy finance systems simply can’t cope. This is no longer a niche IT issue – it’s a frontline finance challenge.”
Along with access via OpenAI, GPT-5 is also available across Microsoft platforms, including Microsoft 365 Copilot, Copilot Studio, Microsoft Copilot, GitHub Copilot, Visual Studio Code and Azure Al Foundry.
Michael Comaous is a dedicated professional with a passion for technology, innovation, and creative problem-solving. Over the years, he has built experience across multiple industries, combining strategic thinking with hands-on expertise to deliver meaningful results. Michael is known for his curiosity, attention to detail, and ability to explain complex topics in a clear and approachable way. Whether he’s working on new projects, writing, or collaborating with others, he brings energy and a forward-thinking mindset to everything he does.
