OpenAI made big promises with the launch of GPT-5 in August, describing it as its “smartest, fastest, most useful model yet, with built-in thinking that puts expert-level intelligence in everyone’s hands”.
Perhaps Sam Altman should have asked it about how to avoid annoying its loyal subscriber base pre-launch, with Plus subscribers – who pay £20 per month – quickly complaining that they had used up their allotted resources.
You now have a choice between Auto, Fast and Thinking, where Fast provides instant answers, and Thinking will take up to 30 seconds to provide a reasoned-through answer. Altman reckoned most people should stick with Auto, which allows GPT-5 to decide how long to take, but paid-for subscribers demanded more control.
At launch, Plus subscribers were limited to 200 Thinking messages per week, fewer than with GPT-4o. OpenAI quickly responded to the complaints, raising the limit to 3,000.
“The upgrades matter: a 400,000-token context window, a set of models that can route work internally for efficiency, and pricing around $1.25 per million input tokens,” said Shanti Greene, head of AI and data strategy at Infocepts. “I have been using it since the Horizon beta. It is fast, consistent, and priced in a way that changes how many teams can justify running larger workloads.”
But smart businesses should make sure they’re using the right models for the right tasks. “This is not about finding the single ‘best’ option any more,” Greene said. “It is about building a toolkit where each model has a defined role. GPT-5 might handle the bulk of general-purpose work. Claude 4.1 Opus is still the safer choice when flawless code matters. Gemini 2.5 Pro is unmatched for processing huge datasets in one shot.”
He added: “That is the new economics in AI. Keep your options open, route work where it makes sense, and use models like GPT-5 to free up budget for the specialised tools that actually move the needle. The winners will be the ones who treat AI like a portfolio to manage, not a single bet to place.”
No comments:
Post a Comment