Thinking Machines Lab releases open-weights Inkling 2 Mini, runnable on a 32GB Mac

Thinking Machines Lab released Inkling 2 Mini on November 20, 2026 under an Apache 2.0 license, making it an open-weights model. The model is built as a looped transformer, reusing its middle layers three times per token, which the lab says adds reasoning compute without many extra written chain-of-thought tokens. Officially supported local use requires Apple silicon with at least 32GB of memory via Ollama, or an NVIDIA DGX Spark, with the checkpoint being about 20GB. The lab reports strong scores on coding and agent benchmarks such as DeepSWE v1.1 and SWE-bench Pro v2 Hard, though the author notes these figures were not independently reproduced. The model supports up to 1M tokens of context and accepts text and image inputs.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in