Skip to content
Computer vision

Inkling 2 Mini

multimodal model with looped architecture

2 min read

Inkling 2 Mini by Thinking Machines Lab is a versatile multimodal model that takes text and image inputs and generates text outputs. It is designed for developers building AI agents, chatbots, coding tools, and Retrieval-Augmented Generation (RAG) systems. The model is built on a 24-layer decoder featuring sparse Mixture-of-Experts (MoE) layers, where middle layers are executed multiple times to boost efficiency. It supports a context window of up to 1 million tokens and excels at complex programming and data analysis tasks.

Key features

  • Multimodal input: support for text and images with resolutions from 40 to 4096 pixels.
  • MoE architecture: 26.4 billion total parameters, with about 3.4 to 5.4 billion active per token thanks to looped layers.
  • Massive context: native support for a context window of up to 1M tokens.
  • Coding support: strong results in automated development and terminal tasks.

What you can do

  • Integrate the model into chatbots and assistants to handle text and visual instructions.
  • ‍Use it for code generation, debugging, and autonomous programming tasks.
  • Set up long-context retrieval-augmented generation (RAG) and information extraction systems.

Example

An example of setting up or running the model using Python libraries or via the Tinker API is available in the official repository and project documentation.

Read moreDetails
Benchmarks: DeepSWE v1.1 — 48.2%, Terminal Bench v2.1 — 71.6%
Hardware requirements: checkpoint around 20GB in NVFP4 and MXFP8 formats; 32GB VRAM or Apple Silicon devices with 32GB+ memory recommended
License: Apache 2.0
Examples
Discuss Telegram 0

Will it run?

Runs in GGUF Q4_K

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • U8weights 26 GB · ready files30 GB
  • FP8 / INT8recommendedweights 25 GB28 GB
  • GGUF Q4_Kweights 15 GB · ready files17 GB
  • 4-bitweights 14 GB16 GB
  • GGUF IQ3_Sweights 12 GB · ready files14 GB
  • GGUF IQ2_Mweights 8.1 GB · ready files9.4 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 5090 · U8 — from RTX 5090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # AMAImedia/Inkling-2-Mini-BF16-GGUF
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download AMAImedia/Inkling-2-Mini-BF16-GGUF --local-dir ~/models/Inkling-2-Mini-BF16-GGUF
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Inkling-2-Mini (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

AMAImedia/Inkling-2-Mini-BF16-GGUF
Author
AMAImedia
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments