Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Inkling 2 Mini

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 5

  1. Transformers v5.24.0 Adds Inkling 2 Mini and Variable-Loop Checkpoint

    Transformers v5.24.0 added support for Inkling 2 Mini, a multimodal model from Thinking Machines Lab that accepts text and image inputs and generates text. According to the release notes, its 24-layer decoder repeats layers 6 through 17 three times, uses a learned message-passing mechanism, and routes each token to 8 of 128 experts with a hidden size of 2304. The release also added the TML-d24-varloop-research checkpoint, which supports a variable number of loop passes, trading speed for quality.

    Transformers Releases · 🔥 10
  2. Artificial Analysis: Inkling 2 Mini Matches Qwen3.8 27B at xhigh

    Thinking Machines released Inkling 2 Mini, an open-weight looped transformer that scored 20 on the Artificial Analysis Intelligence Index. Artificial Analysis says the model matched Qwen3.8 27B at the xhigh reasoning setting while using 5.4 billion active parameters, approximately one-fifth as many.

    The Decoder · 🔥 15
  3. Ollama v0.40.17-rc3 Adds Inkling 2 Mini Support

    The Ollama Releases report lists the v0.40.17-rc3 tag as created on 20 Nov at 18:37. The listed change is “mlxrunner: add Inkling 2 Mini,” adding support for the Inkling 2 Mini model in the MLX runner.

    Ollama Releases · 🔥 8
  4. Inkling 2 Mini launches with day-zero vLLM support

    Thinking Machines announced Inkling 2 Mini, a multimodal mixture-of-experts model with 26B total parameters, 3.5B active parameters and support for contexts up to 1M tokens. Its middle 12 layers run three times per token with shared weights, separate KV caches and gated messages between passes. vLLM Blog says vLLM provides day-zero support across NVIDIA Blackwell, Hopper and DGX Spark, plus serving support for the TML-d24-varloop-research checkpoint.

    vLLM Blog · 🔥 3
  5. Hugging Face ships Inkling 2 Mini, a looped-transformer MoE for local setups

    Hugging Face announced that Thinking Machines Lab's Inkling 2 Mini is available, describing it as a multimodal mixture-of-experts model with about 26B physical total parameters and 3B active parameters per token. Its 24 decoder layers include 12 middle layers that run three times in a loop, giving it the effect of 48 logical layer visits, and it supports a 1M-token context with image and text inputs. Weights ship in mixed precision with NVFP4 and MXFP8 expert weights, a DSpark drafter is included, and support is available in transformers, Llama App, llama.cpp, MLX and SGLang with day-zero inference. Hugging Face says in its informal vibe tests the model was at least comparable with Qwen 3.8 27B, while noting that more validation is needed.

    Hacker News 首页(100+ 分) · 🔥 34

Experience and discussion from the community

Share my Inkling 2 Mini experienceAsk about Inkling 2 Mini

Nobody has shared their experience with Inkling 2 Mini yet.