Thinking Machines Lab released Inkling 2 Mini on November 20, and it is the most interesting local model launch in a while. It is a multimodal open-weights model with 26.4 billion total parameters and only 3.4 billion active per token — and the whole checkpoint is about 20GB, small enough to run on a laptop.
The trick is looping. Inkling 2 Mini has 24 decoder layers in four blocks of six. The first block runs once, the two middle blocks run three times each, and the final block runs once, so each token passes through 48 layer applications even though only 24 layers are stored. It was also trained natively in 4-bit precision — 20 layers in NVFP4, the last four in MXFP8 — and the released checkpoint keeps that exact layout, so there is no conversion loss.
On benchmarks it leads every open-weights model under 100B total parameters on DeepSWE v1.1 (48.2%) and Vals Index v2.2 (46.5%), and it beats Qwen3.8-27B — the model most local coding setups default to — on DeepSWE, SWE-bench Pro v2 Hard, SWE-bench Multilingual, SWE-bench Verified and Vibe Code Bench. Stanford researchers also measured intelligence per joule: Inkling 2 Mini scored 0.703 on an NVIDIA DGX Spark, against 0.430 for Qwen3.6-35B-A3B.
For local inference, the official targets are a single NVIDIA DGX Spark or Apple silicon with 32GB or more of memory. It works with Hugging Face transformers, SGLang, vLLM, llama.cpp GGUF versions, Ollama and MLX, and it is Apache 2.0 licensed, so you can fine-tune it and build it into products. The demos are the fun part: researchers ran it offline on a MacBook Pro in airplane mode during a flight and built a playable flight simulator in a few minutes, and one offline MacBook powered all 20 residents of "Thinkyville", a tiny pixel town.
If you have a 32GB Mac or a DGX Spark-class machine, this is now the local model to beat. The weights are free on Hugging Face, and you can also try it in the Tinker playground without downloading anything.
Read the full article with all the details on Rohan Tech Official: https://rohantechofficial.com/tech-news/inkling-2-mini-local-ai-model/
