← Back to the wire

Meet Thinking Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac

AnnouncementModelNov 21, 2026

Thinking Machines Lab released Inkling 2 Mini, an open-weight looped transformer with 26.4B total parameters that runs its middle layers three times per token. The roughly 20GB checkpoint fits a 32GB Apple Mac or one NVIDIA DGX Spark. Thinking Machines Lab reports 79.2% on SWE-bench Verified and ties Qwen3.8-27B at 20.1 on the Artificial Analysis Index. Weights ship under Apache 2.0 on Hugging Face, with GGUF builds via Unsloth.

Receipt № 30081 source · awaiting confirmation ◐

Evidence

1source· awaiting independent confirmation

No score is assigned. Sources and their independence are shown in the citation chain below.

Citation chain · 1 source

01med
www.marktechpost.comBloghttps://www.marktechpost.com/2026/11/21/thinking-machines-inkling-2-mini-looped-transformer/This source supplies the launch coverage and technical specifics of Inkling 2 Mini, explaining its looped-layer architecture, training pipeline, benchmark results, and supported hardware setups.
PRIMARY
Hugging FaceCompanyNVIDIACompanyUnslothCompanyAppleCompanyThinking Machines LabCompanyQwen3.8-27BModelInkling 2 SmallModelInkling 2 MiniModel
Canonical: https://www.marktechpost.com/2026/11/21/thinking-machines-inkling-2-mini-looped-transformer/