Checked 21 November 2026. Prepared with AI assistance from linked primary sources. No hands-on testing or independent benchmarking. Ordinary source links; no affiliate links.
Inkling 2 Mini: a local option with a meaningful hardware threshold
Thinking Machines Lab released Inkling 2 Mini on 20 November. Its model card lists downloadable weights under Apache 2.0, accepts text and image inputs, and produces text. The checkpoint is about 20GB.
Access and restrictions: the vendor lists Apple Silicon devices with at least 32GB memory among supported local setups. It recommends at least 32GB VRAM for other GPU platforms. Consult the model card’s acceptable-use policy and software requirements before deployment. “Mini” should not be read as a promise that every office laptop can run it.
Tecrave analysis: a small studio with suitable existing hardware and technical help could evaluate local drafting or image-description tasks. Compare outputs with source material before using them in client work. Downloadable weights do not eliminate equipment, electricity, maintenance or review costs.
Hugging Face provides local-runtime guidance, including MLX and llama.cpp support. Its performance assessments are publisher claims, not Tecrave validation. Start with non-sensitive examples and confirm whether the chosen application sends anything to external services before assuming a private workflow.
Original model card and release date; Hugging Face launch and runtime guidance, 20 November.