Supporting case
Thinking Machines releases Inkling 2 Mini for local deployment
What happened
Thinking Machines released Inkling 2 Mini on November 20 with open weights listed under Apache 2.0 ↗. Its 26.4 billion parameters include 3.4 billion activated for a token; repeated passes through middle layers perform additional computation without storing a correspondingly larger model. The company reports an approximately 20 GB memory footprint and documents local deployment routes. Its offline MacBook demonstration and performance claims are company evidence, not independent device tests.
A smaller locally deployable model can reduce reliance on a remote inference provider. Memory, runtime support, useful performance and the surrounding tools remain dependencies; downloadable weights do not by themselves replace a hosted service.
Dates and implementation stage
Action 2026-11-20; publication 2026-11-20; verified/approved 2026-11-21
