Ling-3.1-flash surfaces: ~560B total params, ~25B active per token, up to 1M context

Muhlwa_Sholanke · reddit · 2026-10-01

A Reddit post claims specs for Ling-3.1-flash: a 560B-parameter MoE model with 25B active parameters per token and up to a 1M-token context window. It reportedly offers a two-week free trial before being open sourced. Details are unverified pending official confirmation.

Related event: Ant's Ling-3.1-flash: 560B Parameters, 1M Context(3 posts)→

Original post →

More from Models

Models channel →