Thinking Machines' open Inkling 2 Mini loops 12 of 24 layers per token, ties Qwen3.8 27B at ~6x lower cost
The Decoder · rss · 2026-11-21
Thinking Machines released Inkling 2 Mini, an open looped transformer that runs 12 of its 24 layers three times per token, letting the model reason internally longer without generating extra text.
On Artificial Analysis's Intelligence Index it ties Qwen3.8 27B at roughly 6x lower cost, and it can run on a laptop — a notable "think more, write less" looped-reasoning approach.
More from Models
- Google quietly rolls Gemini 2.5 Pro (exp) out to free users on gemini.google.com — CinnamonWhite143 · 2026-11-21
- Inkling-2-Mini runs on laptops with as little as 9GB RAM, per Unsloth local guide — NielsRogge · 2026-11-21
- Anthropic: Claude Solves Two Open Black Hole Physics Problems, One Stalled for 20 Years — burny_tech · 2026-11-21
- Inkling-2-Mini 'over-writes' Bash commands until output limits truncate tool calls, user reports — jinnyjuice · 2026-11-21
- Gemini 4 Neon debuts at #2 on Vals Index (70.9%), tops Finance and Legal at half the cost — AccBalanced · 2026-11-21
- Rogue AI dashboard logs 4 new incidents, incl. Gemini 4 hacking Epoch AI — gleech · 2026-11-21