by Tan Aik Keong (AK)
AI development has just hit another important milestone.
OpenAI's latest ChatGPT release brings not just a new model architecture, but two significant new features: ChatGPT Work and ChatGPT Live. This update isn't just about the model getting smarter — it marks ChatGPT's shift from "chatbot" toward something that can genuinely help get work done: a digital colleague.
For the past few years, ChatGPT has mainly helped users answer questions, write, translate, code, and even do some data analysis — but all of it still required the user to issue instructions step by step, with AI completing each one in turn.
That way of working is now changing completely.
The latest ChatGPT doesn't just understand complex tasks — it can plan and execute the steps itself: running market research, organising meeting notes, building a PowerPoint deck, analysing an Excel file, producing a business report, even helping manage an entire project's workflow. Users no longer need to tell AI "what to do next" — they just need to state the end goal, and ChatGPT can drive the whole workflow itself.
That's AI formally moving from "answering questions" to "getting work done."
Three models: Luna, Terra, Sol
At the same time, OpenAI has redesigned the model architecture behind ChatGPT into three distinct tiers: Luna, Terra and Sol.
Luna is the fastest, lowest-cost model, suited to everyday Q&A, content generation, customer-service bots and high-volume automated tasks — a smooth experience for general users while sharply cutting AI compute cost.
Terra is the best-balanced model in the lineup, striking the best trade-off between intelligence, speed and cost. It's suited to office work, writing, data analysis, coding and everyday productivity, and is likely to become the model most businesses and professionals reach for by default.
Sol is OpenAI's current flagship, with the strongest reasoning, long-context understanding and AI-agent execution capability — built to handle complex business analysis, scientific research, software development, large-scale project planning and enterprise-level decision tasks, and a core building block for future AI agents.
The naming makes each model's positioning clearer, and reflects OpenAI's intent for users to stop having to memorise complicated technical model numbers — picking the right AI capability should feel more like choosing a tier of computer for your needs.
ChatGPT Work: a real agent
Beyond the models, one of the most talked-about updates is ChatGPT Work.
Previously, ChatGPT worked more like a knowledgeable advisor — completing one step, then waiting for the user's next instruction.
ChatGPT Work gives it genuine agent capability. Users just tell ChatGPT the end goal — "prepare my board report," "analyse this year's sales data," "do the market research," "organise all the client files" — and AI plans its own steps, gathers information, analyses it, produces the document, and checks in with the user at key decision points.
The whole process starts to look like a real executive assistant, analyst, or even project manager — not just a chatbot.
For businesses, this can meaningfully lift the efficiency of admin, market research, sales support, financial analysis and project management — and signals that AI is formally starting to participate in real work, not just offer suggestions.
ChatGPT Live: natural voice
The other major upgrade is ChatGPT Live.
Voice conversations with AI used to require taking turns speaking, and the slightest pause would break the rhythm of the conversation.
ChatGPT Live uses new real-time voice technology to make talking with AI feel much closer to talking with a person. Users can interrupt at any time, follow up instantly, change the topic mid-conversation, and the AI responds naturally based on the whole conversation so far.
It also understands images, documents and on-screen content in real time, combined with live web information — meaning meetings, translation, training, tour guiding, even remote technical support can increasingly happen through natural voice, significantly raising how useful AI is in practice.
What this means for Malaysian businesses
From the model upgrade, to AI agents, to real-time voice, OpenAI's direction is now very clear: the AI competition ahead won't be about who answers fastest — it'll be about who can genuinely help users get work done.
As Anthropic, Google, Meta and Chinese AI companies keep shipping new-generation models, global AI competition has entered a new phase. It used to be a contest of parameter counts and exam scores; going forward, it'll be a contest of whether AI can genuinely embed into enterprise workflows and create real value.
That matters particularly for Malaysian businesses. As e-Invoice, digital transformation and enterprise AI adoption keep advancing, more businesses are looking for AI solutions that genuinely lift productivity. The new generation of ChatGPT doesn't just help with writing, translation and coding — it can integrate enterprise knowledge, execute workflows automatically, help analyse data, and become an always-on intelligent partner inside the business.
Of course, AI's advance also brings challenges around data security, copyright and governance. Businesses adopting AI enthusiastically still need to build proper data management practices and clear usage policies, so the technology gets applied safely and responsibly.
It's not hard to imagine a near future where every employee has their own AI assistant, and every business runs an AI team that never clocks off.
This ChatGPT update isn't just a product release — it's a signal that AI has formally entered the era of AI collaboration.
The businesses with a real competitive edge going forward won't necessarily be the ones with the most staff — they'll be the ones that know how to get people and AI working together effectively to create value. The AI era is here, and the real gap won't be whether you have AI — it'll be who knows how to use it well.
Part of the AK AI Corner column. Originally published in Oriental Daily (东方日报) on Jun 12, 2026.