DigiNews

Tech Watch by Johan Denoyer

← Back to articles

Power consumption of LLM's

Quality: 7/10 Relevance: 9/10

Summary

A community discussion on the power consumption of large language models, questioning whether major vendors publish per-token energy metrics and training costs. The thread cites a GPT-OSS energy post and a JP Morgan data-center study, discusses rough energy per token estimates, and debates context size, efficiency, and environmental impact.

🚀 Service construit par Johan Denoyer