Scan to Download Gate App
qrCode
More Download Options
Don't remind me again today

Noticed something wild lately - certain AI models like Sonnet 4.5 can apparently keep running for over 30 hours straight, and Codex isn't far behind with its extended operation windows. Got me wondering about the architecture behind this.



Anyone come across research papers or technical docs that dig into how these systems sustain such long inference sessions? Curious whether it's about model architecture innovations, infrastructure optimization, or something entirely different. Would love to see what the research community is saying about this capability.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 5
  • Repost
  • Share
Comment
0/400
NewDAOdreamervip
· 9h ago
30 hours? Come on... How much would that burn the card? Has anyone really used it like this?
View OriginalReply0
TaxEvadervip
· 9h ago
30 hours of continuous operation? How much electricity bill will that cost? That's a bit ridiculous.
View OriginalReply0
SchroedingerAirdropvip
· 9h ago
30 hours without dropping the connection? How much electricity would that consume?
View OriginalReply0
GasFeeCryingvip
· 9h ago
30 hours of no downtime, this number is ridiculous, right? Is it true or not?
View OriginalReply0
GasGasGasBrovip
· 10h ago
30 hours of non-stop running? Is this burning video memory or some kind of black magic?
View OriginalReply0
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate App
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)