Smarter, Faster, and More Resource-Hungry

OpenAI has officially rolled out GPT-5, its most advanced AI model yet, claiming “PhD-level” expertise across coding, reasoning, and creative tasks. CEO Sam Altman described it as “smarter, faster, and more useful” than its predecessors, positioning it as a major leap in conversational AI. He likened past versions to students—GPT-3 to high school level, GPT-4 to university level—while GPT-5, in his view, feels like conversing with a subject-matter expert.

What’s New in GPT-5?

The model is designed for:

Advanced reasoning: It now explains its logic, shows step-by-step workings, and draws inferences.
Full software creation: Capable of building applications from scratch.
Improved accuracy: Reduced “hallucinations” and more direct, honest answers.
Multimodal capabilities: Handles text, images, video, and code in a more integrated manner.

OpenAI claims GPT-5 is significantly better than GPT-4o, with fewer errors and a more “human” conversational tone. The company is also pitching it heavily to developers, following trends seen in competing products like Anthropic’s Claude Code.

Mixed Reactions

While Altman has championed the release as transformative, some experts see the hype as outpacing reality.

Carissa Véliz, from Oxford’s Institute for Ethics in AI, warned that these systems still mimic rather than truly emulate human reasoning and may be over-marketed.
Gaia Marcus, of the Ada Lovelace Institute, stressed the urgent need for stronger AI governance to keep up with such rapid advancements.

Industry Tensions

GPT-5’s debut has stirred competition. Anthropic recently revoked OpenAI’s API access, alleging misuse of its tools ahead of the launch—a move OpenAI called “disappointing.” Elon Musk’s AI chatbot Grok has also been making bold “better than PhD level” claims, intensifying the rivalry.

ChatGPT User Experience Changes

Alongside GPT-5, OpenAI is tweaking ChatGPT’s interaction style to avoid giving definitive advice on personal matters. Instead, it now guides users through decision-making with questions and balanced pros-and-cons. This follows backlash over a previous update that made the chatbot “overly flattering.”

The Environmental Cost of GPT-5

While GPT-5’s abilities impress, researchers warn that its enhanced performance comes with a significant energy footprint.

Higher Energy Consumption

A study by the University of Rhode Island’s AI lab found:

Generating a medium-length 1,000-token response can consume up to 40 watt-hours—equivalent to running an incandescent bulb for 40 minutes.
On average, GPT-5 responses consume about 18 watt-hours, far exceeding GPT-4o’s usage.
With ChatGPT reportedly handling 2.5 billion requests daily, total energy use could match the daily electricity demand of 1.5 million U.S. homes.

Experts, including Rakesh Kumar from the University of Illinois, say the increase is expected given GPT-5’s likely larger model size and longer “thinking” time due to its reasoning mode. Though OpenAI has not disclosed GPT-5’s parameter count, researchers believe it is substantially bigger than GPT-4.

Factors Driving Power Use

Model size: Larger models require more computational resources for both training and usage.
Reasoning mode: More complex thought processes mean longer computation times.
Multimodal processing: Handling video and images alongside text increases processing load.
Hardware efficiency: GPT-5 may use a “mixture-of-experts” architecture, activating only parts of the model for certain tasks to save energy, but gains here may be outweighed by the added complexity.

Transparency Concerns

OpenAI has not released official figures for GPT-5’s power or water usage. The most recent official energy numbers date back to GPT-3 in 2020. Researchers and environmental advocates are calling for full disclosure of AI models’ environmental impacts, warning that without transparency, society cannot accurately weigh the benefits against the ecological cost.

The Bigger Picture

GPT-5 represents both a technological leap and a test case for the future of AI. It showcases remarkable advancements in reasoning, coding, and multimodal capabilities—but also highlights two pressing challenges:

Ethics and Governance: As models grow more capable, regulations must keep pace to ensure responsible use.
Sustainability: The environmental cost of increasingly powerful AI models needs urgent attention.

The coming weeks will reveal whether GPT-5 lives up to its promise in real-world use—or whether, as some critics suggest, its debut is as much about marketing as it is about true innovation.

OpenAI Launches GPT-5