Anti-woke Grok is actually the most environmentally friendly chatbot

Despite some skepticism about climate change among the “anti-woke” crowd, Elon Musk’s chatbot Grok turns out to be the least harmful to the environment – while ChatGPT ranks as the worst.

Grok is the most environmentally friendly artificial intelligence (AI) chatbot, emitting just 0.17 grams of CO₂ per query, according to TRG Datacenters, a data infrastructure provider.

The model’s architecture appears to be specifically designed to reduce power consumption, with each query using roughly the same amount of energy as a single Google search, the firm said. This makes Grok a “standout performer” among major chatbots.

ChatGPT, its much larger rival from OpenAI, ranks at the opposite end of the spectrum. The GPT-4 model produces 25 times more carbon emissions per query than Grok, at 4.32 grams of CO₂ per request.

A single ChatGPT query equates to sending 21 emails or nearly a full phone charge in terms of the carbon emissions it generates.

“As AI adoption continues to rise, finding ways to reduce its energy consumption will be key. Some models are already designed to be more efficient, but there is still room for improvement,” a spokesperson from TRG Datacenters said.

Advances in hardware, more optimized AI models, and increased use of renewable energy in data centers could help lower emissions over time.

“AI is here to stay, but balancing innovation with sustainability will be essential in minimizing its environmental impact,” the firm said.

Billionaire X owner and Tesla CEO Elon Musk has positioned Grok as an alternative to “woke” chatbots like ChatGPT, even though some studies suggest that all major AI models – including Grok – tend to lean left.

The study also evaluated Google Gemini, which ranked as the second most energy-efficient chatbot after Grok, followed by Meta’s LLaMA, Claude AI, and Perplexity AI – with the latter being nearly as energy-hungry as ChatGPT.

According to researchers, the environmental impact of AI models also correlates with their computational complexity, with search–integrated models such as Perplexity generally showing higher emissions.

The company calculated carbon emissions based on standard energy grid assumptions and compared them to everyday digital activities such as sending emails, streaming video, and charging smartphones.

For example, Gemini’s 1.6 grams of CO₂ emissions per query is equivalent to watching a 10-minute YouTube video. Meta’s model emits 3.2 grams per query – roughly the same as sending 10 simple emails.

Claude’s 3.5 grams is comparable to watching a 10-minute YouTube video and sending an email, while Perplexity’s 4 grams is roughly equal to charging a smartphone one to one and a half times.