Inkling Small (free)

Thinking Machines· released 30 Jul 2026

thinkingmachines/inkling-small:free
Input / 1M tokens
free
Output / 1M tokens
free
Context window1M tokens

Longest answer: 262K tokens

reads textreads imagereads audioopen weightstool callingvisionreasoning

About

Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines. It has 276B total parameters, of which 12B are active, and is described as the smaller, more efficient member of its model line. It accepts text, image, and audio inputs and returns text. Listed capabilities include tools, vision, and reasoning. Its context window is 1,048,576 tokens, with a maximum output of 262,144 tokens. The model is free and its weights are open; the Hugging Face ID is `thinkingmachines/Inkling-Small`. Its model ID is `thinkingmachines/inkling-small:free`, and it was released on 2026-07-30. For teams evaluating a multimodal model, the combination of open weights, listed input types, and large context and output limits defines its stated scope.

Who it is for

It suits teams seeking an open-weight model for text, image, and audio inputs, with tool use and reasoning listed. The free listing and large context and output limits may fit evaluation of longer multimodal workflows.

What is good

  • Free listing.
  • Open weights.
  • Accepts text, image, and audio inputs.
  • Supports tools, vision, and reasoning.
  • 1,048,576-token context window.

What to know first

  • Returns text only.
  • Maximum output is 262,144 tokens.

Verdict

Inkling Small combines open weights with text, image, and audio inputs and text output. Its parameter counts and listed limits offer concrete points for assessing fit.

Details

Lab
Thinking Machinesopenrouter.ai · 3 Oct 2026
Context
1,049Kopenrouter.ai · 3 Oct 2026
Input price
Freeopenrouter.ai · 3 Oct 2026
Output price
Freeopenrouter.ai · 3 Oct 2026
Max output
262,144 tokensopenrouter.ai · 3 Oct 2026
Inputs
text, image, audioopenrouter.ai · 3 Oct 2026
Open weights
Yesopenrouter.ai · 3 Oct 2026

More longest-context models

See the list