Back to articles
DevPickView feed
Hacker NewsHN submitter: Argonautlabs
Original published:

Kimi K3 (2.8T) at 1 token/s on a MacBook Pro, streamed from four SSDs

AI summary
Editorial policy

MacBook Pro上で複数のSSDを使用してKimi K3モデルをストリーミングし、1トークン/秒の速度で実行する技術的な実証が行われた。

Key points

  • •2.8Tパラメータを持つKimi K3モデルをMacBook Proで動作させる試みが報告されている
  • •4つのSSDからモデルをストリーミングすることで推論を実現した
  • •1トークン/秒という低速だが、エッジデバイスでの大規模モデル実行の可能性を示す実験である
#Semiconductors & Infrastructure#LLM / Model Trends
Read the original(opens in a new tab)

Related articles(Semiconductors & Infrastructure)

See more
Hacker News

Why harnesses matter for energy (and token) efficiency

Analysis of a month of agent traffic reveals that the harness, not just the model, determines the energy consumption of coding sessions.

The Verge

A week with RTX Spark: everything we’ve learned using Nvidia’s new laptops

A review article testing the performance and usability of laptops equipped with Nvidia's Arm-based SoC, the RTX Spark N1X.

Hacker News

Apple ML Research: Muse Spark Open with MLX on Mac Studio

An introduction to using the MLX framework on Apple silicon Macs, allowing AI researchers and developers to experiment locally with the latest models.

Hacker News

Qualcomm Legally Proves Arm Breached Its License, Wins $2

A Delaware jury found that Arm breached its Architecture License Agreement with Qualcomm, but Qualcomm was awarded only $2 in nominal damages.

PC Watch

Hothot Review: Running Large Local AI with 128GB Unified Memory on the RTX Spark-Powered ProArt P14

The ASUS ProArt P14, a 14-inch laptop featuring NVIDIA's new RTX Spark platform, launches on October 23.