Thinking Machines releases Inkling 2 Mini: looped-transformer open model that runs on 20GB locally
mark_k · x · 2026-11-21
Thinking Machines Lab released Inkling 2 Mini, an open-weights model that runs locally on a laptop.
- Key architectural twist: a looped transformer that runs its middle layers three times per token, doing more internal reasoning without emitting extra reasoning tokens — particularly efficient for agentic coding
- 26.4B total params (3.4B active), needs only 20GB of memory; per official numbers it beats every open-weights model under 100B on several coding/agentic benchmarks (e.g. 77.9% on SWE-bench Pro v2 Hard, 79.2% on SWE-bench Verified)
- Demo: built a playable flight simulator in minutes, fully offline on a MacBook Pro
- Weights are available now
A notable release for the local-AI ecosystem.
More from Models
- Downdetector reports OpenAI outage starting 3:13 PM EST as user complaints pile up — VoidStateKate · 2026-11-21
- Inkling 2 Mini on Ollama: MLX build, one-command hooks into Claude Code and OpenCode — ollama · 2026-11-21
- Inkling 2 Mini demoed building an RTX 5090 scene in Three.js — ollama · 2026-11-21
- MaxxBench cost-quality chart: Muse Ember 1.0 hits 63.5% at $0.18, crushing rivals on value — AIatMeta · 2026-11-21
- Thinking Machines' Inkling-2-Mini trends on Hugging Face — thinkingmachines · 2026-11-21
- dottxt tool-call index: DeepSeek tops 100M daily calls, failure rates vary 0.04%–17% by provider — remilouf · 2026-11-21