# TML releases Inkling 2 Mini, an open-weights 26.4B-parameter transformer that loops middle layers three times per token for better reasoning with fewer tokens (Thinking Machines Lab)

> HelixNews · Web page: https://helixnews.one/story/11323622 · All pages for agents: https://helixnews.one/llms.txt

3 outlets covering · First listed: Techmeme, Nov 20, 2026, 3:15 PM · Read it: <https://www.techmeme.com/261120/p33>

## Who's covering it

In the order they published:

- Nov 20, 2026, 2:12 PM: [Inkling 2 Mini](<https://thinkingmachines.ai/news/inkling-2-mini/>) — Hacker News (Aggregator, United States)
- Nov 20, 2026, 2:28 PM: [Thinking Machines Releases Inkling 2 Mini for Local AI](<https://thedefiant.io/frontier/ai/thinking-machines-releases-inkling-2-mini-for-local-ai>) — The Defiant (Specialist)
  The roughly 20GB model supports Apple silicon devices with 32GB or more memory. Its Apache 2.0 listing comes with a separate acceptable-use policy.
- Nov 20, 2026, 3:15 PM: [TML releases Inkling 2 Mini, an open-weights 26.4B-parameter transformer that loops middle layers three times per token for better reasoning with fewer tokens (Thinking Machines Lab)](<https://www.techmeme.com/261120/p33>) — Techmeme (Aggregator, United States)

## How outlets word it

Most outlets use: Inkling, Mini, Machines, releases, Thinking.

- Only Techmeme says: TML, open-weights, 4B-parameter, transformer, loops, middle, layers, three, times, token, better, reasoning, fewer, tokens, Lab
- Only The Defiant says: Local
