GitSeonjeap
English

← News

Local AI

Inkling 2 Small Open Weights Released, 594B MoE Model

Posted As of

Thinking Machines Lab has released the weights for Inkling 2 Small. The model card lists Apache-2.0 and a 1M token context, but one analysis pointed out that the repository has no LICENSE file.

What happened

  • Based on the Thinking Machines Lab announcement and the Hugging Face model card, the weights were released on 2026-11-12.
  • The initial launch was on 2026-11-05, and at that time it was hosted only on Tinker.
  • The content of the November 5 announcement was confirmed through CASRAI's summary, and the intent is to release under Apache 2.0 after external safety testing and red-teaming.
  • (Official) According to the model card, it takes text and images as input and outputs text, with a maximum context of 1M tokens.
  • (Official) It is a sparse MoE with 17B parameters active per token out of 594B total parameters. There are two repositories, Inkling-2-Small and MXFP8, and both are listed as 594B.
  • (Third-party material) The Artificial Analysis Intelligence Index score is 31.

Our take

The key question for running it locally on a Mac or PC is how to load weights at the 594B scale, and hardware requirements and self-hosting conditions have not been confirmed from a primary source. There is also a point that the metadata and the presence of a LICENSE file differ, so the repository should be checked directly before use. The benchmark figure comes from third-party material, so it is hard to trust as is.

Sources and related posts

More in Local AI

  1. NVIDIA Releases Nemotron for Science for Scientific Agents (nvidianews.nvidia.com)

    NVIDIA announced a package based on Nemotron 3.5 Super VL. It includes a chemistry dataset and a post-training recipe, and the IR copy states that the model is optimized to run locally on a 128GB DGX Spark.

    Local AI| |1 sources |Our take included
  2. NVIDIA RTX Spark Compact Desktops Begin Shipping (blogs.nvidia.com)

    Compact RTX Spark desktops from ASUS, Dell, and MSI have launched as of the announcement date and are positioned for local AI agents and model fine-tuning. Pricing was not confirmed on NVIDIA's page.

    Local AI| |1 sources |Our take included
  3. llama.cpp b12497 adds tiny-voca support (github.com)

    llama.cpp pre-release b12497 adds tiny-voca support to the multimodal path and improves G2P conversion. The release page has no performance figures or description of the target model.

    Local AI| |1 sources |Our take included

All news