DeepSeek V3 inference

What experts have said about DeepSeek V3 inference

1 statement · 1 positive

1. Dwarkesh PatelPositiveAug 7, 2026· Dwarkesh Podcast

   Sparse-model inference may be most efficient above 2,400 concurrently generated sequences.

   “the optimal inference batch size for a sparse model like say, deep seq v3 is more than 2,400 concurrent sequences being generated at once.”

   Listen at 7:22

   Open the episode · 8 Predictions for the Era of Continual Learning

Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.