PulseAugur
LIVE 12:29:17
tool · [1 source] ·
0
tool

Gemini Ultra training consumes 150 GWh; Nano version uses negligible energy

A Mastodon user shared energy consumption figures for Google's Gemini models, estimating that training Gemini Ultra required approximately 150 GWh. Distilling a smaller 'Nano' version from Ultra is estimated to consume between 1 and 5 GWh. A single Gemini cloud query uses about 0.24 Wh, while a year of serving Gemini (around 1 billion queries daily) consumes roughly 90 GWh. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Provides insight into the significant energy costs associated with training and running large AI models.

RANK_REASON The cluster contains a user-generated estimate of energy consumption for AI models, not an official release or benchmark. [lever_c_demoted from research: ic=1 ai=0.7]

Read on Mastodon — sigmoid.social →

COVERAGE [1]

  1. Mastodon — sigmoid.social TIER_1 · [email protected] ·

    @ bhg Thanks for sharing your thoughts. Here are the actual numbers Train Gemini Ultra ~150 GWh Distill Nano from Ultra ~1–5 GWh / ~1–3% One Gemini cloud query

    @ bhg Thanks for sharing your thoughts. Here are the actual numbers Train Gemini Ultra ~150 GWh Distill Nano from Ultra ~1–5 GWh / ~1–3% One Gemini cloud query 0.24 Wh / 0.00000016% 1 year of Gemini cloud serving (~1B queries/day) ~90 GWh / ~60% 👉One Nano on-device query ~0 datac…