Inkling 2 Mini: Thinking Machines' looped open model that runs on a laptop
A 26.4B-parameter open-weights MoE that loops its middle layers three times per token, built to run agentic work on a laptop.
Why it matters
Frontier-lab reasoning tricks (looped transformer) in a ~20GB model you can run locally instead of renting cloud GPUs.
What you could build with it
A small startup could ship a fully offline coding assistant inside its desktop app, running this looped model on a laptop GPU so developers get agentic code help on planes and in air-gapped offices without any API bill.
Does it hold up?
Too early to judge: released 20 Nov and the first independent local runs have not appeared yet; early benchmarks put it ~6x cheaper and ~4x faster per task than Qwen3.8 27B.
Built with Inkling 2 Mini
- Thinking Machines releases Inkling 2 Mini open-weight modelarticle · Writeup of the looped-transformer design, the 26.4B/3.4B-active sizing, and Artificial Analysis figures putting it ~6x cheaper and ~4x faster per task than Qwen3.8 27B.
- Thinking Machines Lab releases Inkling 2 Mini for local devicesarticle · Coverage of the laptop-scale pitch, Apache 2.0 open weights, and availability on Hugging Face and Tinker.
- Inkling 2 Mini, an open looped model that runs on a laptoparticle · Daily brief noting the 20 November release, Hugging Face weights, and Tinker inference and fine-tuning.
Learn more
First spotted on article: source.
More AI models
MiniMind: train your own 64M LLM from scratch in about 2 hoursA complete, from-scratch LLM training pipeline at 64M parameters - pretraining, SFT, LoRA, DPO, PPO/GRPO, tool use and…models · JEV 0.82VoiceStudio: the open-source, fully-local ElevenLabs alternativeA desktop app bundling 20 local speech-generation engines and 10 transcription options behind one UI — voice cloning…models · JEV 0.76Edge0-145B: open 145B MoE that runs on a single consumer GPUEdge0 open-sourced 145B and 120B latent-MoE models derived from Qwen3.8-Flash-Next, plus an inference framework that…models · JEV 0.74Strata: run a 125B model on a gaming PCFree, MIT-licensed local inference engine that runs Qwen3.8-Flash-Next (125B MoE, about 6B active per token) on…models · JEV 0.73tiny-voca: a fully open 4.5M-parameter TTS trained for under $10 of computeA fully open 4.5M-parameter English TTS — code, weights and dataset under Apache 2.0 — trained from scratch in 7.7…models · JEV 0.67Qwen4-Max: Alibaba's new flagship model goes liveAlibaba's 2.4T-parameter flagship is now generally available via API, with a 1M-token context, native vision, and…models · JEV 0.67
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.