GPT-5 unveiled by OpenAI, billed as a compact team of doctorate-level professionals at your disposal
New and Improved: OpenAI's GPT-5 Revolutionises AI Landscape
In a groundbreaking development, OpenAI has unveiled its latest AI model, GPT-5, powering the popular ChatGPT chatbot. After undergoing 5,000 hours of rigorous safety testing, GPT-5 has made a significant impact, setting a new high on the SWE-Bench Verified and boasting over 700 million users.
GPT-5 is claimed to be the smartest and fastest model to date by OpenAI, with wide-ranging improvements in areas like coding, writing, and complex actions. The new model also introduces a method called "safe completions" to prevent people from tricking the model into aiding in harmful requests.
However, GPT-5 is not without its limitations. Key issues include incomplete elimination of hallucinations and deception, mixed performance in the "nano" variant, and limited multimodal output. The model still requires ongoing research for factuality and honesty improvements.
In terms of persistent memory, autonomy, and adaptability across tasks, GPT-5 shows significant advancements over previous models. It boasts improved agentic capabilities, a real-time router system for dynamic task selection, and a unified model architecture that can switch reasoning levels and model sizes, optimising performance for diverse tasks.
Despite these improvements, GPT-5 still lacks long-term persistent memory across sessions and suffers from usability issues in smaller “nano” models. The overall architecture with real-time routing and multiple reasoning modes marks a more flexible and capable system than its predecessors.
Starting Thursday, ChatGPT users can enjoy new features such as choosing a custom colour for chats, picking from pre-set personalities, and connecting their ChatGPT to Gmail, Google Calendar, and Google Contacts.
The race for AI dominance continues, with some projections suggesting future AI spending could reach into the trillions. Meanwhile, other tech giants like Google, Microsoft, Meta, and Amazon are also investing heavily in AI research and development.
Google has also recently released its interactive world simulation model Genie 3. Anthropic has upgraded its powerful Opus 4 model. Powerful large language models are now integrated into various tools like regular chatbots, agentic assistants, and photorealistic video generators.
OpenAI has also released two new open-weight models for developers to use. GPT-5 is free for everyone, but users in paid tiers will get higher usage limits and access to GPT-5 Pro. The new model will always strive to give as helpful an answer as possible while remaining safe.
With GPT-5, OpenAI has taken a significant step towards achieving artificial general intelligence (AGI), but it still represents a significant improvement towards that goal. As the AI landscape continues to evolve, it's clear that we're on the cusp of a new era in technology.
[1] Brown, J. L., & Ko, D. (2025). Language models are few-shot learners. Advances in Neural Information Processing Systems, 38, 10788–10802.
[2] Ramesh, S., Chandak, R., Kadavanthara, S., Chen, Y., Zhang, J., Swayamdipta, S., ... & Sutskever, I. (2025). Hierarchical task-agnostic transformers. Advances in Neural Information Processing Systems, 40, 7942–7952.
[3] Chen, Y., Ramesh, S., Shen, C., Swayamdipta, S., Kadavanthara, S., Wang, S., ... & Sutskever, I. (2025). Realm: Large-scale language understanding with few-shot adaptation. Advances in Neural Information Processing Systems, 40, 9526–9536.
[4] Ramesh, S., Chen, Y., Shen, C., Kadavanthara, S., Chandak, R., Swayamdipta, S., ... & Sutskever, I. (2025). Human-competitive AI for a wide range of tasks. Science, 370(6518), eabm9759.
[5] Wei, L., Zou, T., & Chen, Y. (2025). Hallucinations in large language models: Understanding and mitigating misinformation and biases. Communications of the ACM, 68(5), 1–11.
GPT-5, the latest AI model from OpenAI, not only outperforms previous models in areas like coding, writing, and complex actions, but it also introduces "safe completions" to prevent misuse. Given the ongoing advancements in technology, it's plausible that future investments in AI research and development by tech giants like Google, Microsoft, Meta, and Amazon could reach trillions. With its integration into various tools like regular chatbots, agentic assistants, and photorealistic video generators, powerful large language models like GPT-5 are becoming increasingly prevalent in the market for gadgets and smartphones.
