← Back to Model Beat
Models·3d ago·all news from November 18, 2026

Why your fine-tuning job takes longer than the FLOPs say

Fine-tuning AI models often takes significantly longer than theoretical computational estimates predict due to overhead from data movement and memory bottlenecks. While researchers typically measure performance by raw floating-point operations, real-world training efficiency is frequently limited by the speed at which hardware can transfer data between memory and processors. Addressing these hidden costs requires optimizing data pipelines and memory allocation rather than simply increasing raw compute power.

Covered by 1 source

Related stories

ModelsAlibaba Unveils New Flagship AI Model in Riposte to US RivalsNov 16 · 28 sourcesModelsAnthropic Researcher Claims to Solve Riemann Problem With ClaudeNov 20 · 17 sourcesModelsIntroducing Gemini 4 NeonNov 18 · 8 sourcesModelsGoogle’s Gemini Can Now Control Mac Devices on Users’ BehalfNov 19 · 4 sources