stealthmodels.

MODEL BENCHMARKS

Inkling-Small: Newton's Mirror

Thinking Machines’ smaller open model activates 12B of its 276B parameters per token and takes text, images and audio across a million-token context.

Explore the results Specs and pricing
1 model1 resultsHow StealthMark works
StealthMark27.6/100Overall benchmark scoreGeneral Intelligence25.5/100Spatial & physical reasoningVisual Intelligence31.1/100Visual construction & detail

DATASHEET

Inkling-Small specs

A good fit forA base for fine-tuning on Tinker, and lower-cost multimodal agent work over long documents, images and audio.

Inkling-SmallOther modelsMedian

Context windowTokens in one request
1,048,576Joint largest with 20 others
Developer
Thinking Machines Lab
Family
Inkling
License
Apache 2.0
Open model
Yes
Local model
No
Parameters
276B
Active parameters
12B
Input → output
Text, image, audio → text
Architecture
42-layer decoder-only transformer with sparse mixture-of-experts feed-forward (6 of 256 routed experts plus 2 shared per token); hybrid local and global attention; natively multimodal with a hierarchical patch image encoder and discrete-token audio

Self-host Inkling-Small

Hardware
Multi-GPU serverEstimated
Download size
74.8 GB (GGUF UD-IQ1_S) to 532 GB (BF16)

Official BF16 weightsOfficial NVFP4 weightsUnsloth GGUF weights

All prices and providers
Deployment details
  • Official BF16 safetensors (thinkingmachines/Inkling-Small)532 GB · BF16
  • Official NVFP4 safetensors (thinkingmachines/Inkling-Small-NVFP4)171 GB · NVFP4, mixed (packed FP4 with F8_E4M3 and BF16 modules)
  • Unsloth GGUF UD-IQ1_S (unsloth/Inkling-Small-GGUF)74.8 GB · GGUF UD-IQ1_S
Measurements and sources8

How StealthMark scores models · General and Visual Intelligence

StealthMark score

StealthMark measures how well a model’s abilities carry over to original, demanding tasks it could not have trained for: physical and spatial reasoning, planning across many steps, and precise visual work. How StealthMark works explains every benchmark, what it tests and why it is hard.

General Intelligence

The ability to reason, plan, solve problems and understand how things relate and behave. A high score means it thinks hard problems through, from careful analysis to plans that span many steps.

Visual Intelligence

The ability to understand visual intent and turn ideas, descriptions and plans into coherent, effective visuals. A high score means design ideas come out the way you meant them, in illustrations, icons, diagrams or interface visuals.

Overall score

Overall shows how well a model’s abilities carry over to original, demanding tasks, weighting every benchmark into one score from 0 to 100. Estimated means the total includes an estimate for one missing scene.

Judging

Scores reflect how well each result meets the benchmark requirements. Game scores come from the recorded play: the points earned on each level and how each level ended. How StealthMark keeps scores honest.

Games

Each game is a sequence of levels, and earlier moves carry into later levels. A level that was not played is shown as not played and never counts as zero. Play every level of Latch or Foresight yourself to get a score calculated like the models’.

Reasoning efficiency

Good results with little thinking score high. On each task, a model’s result and reasoning tokens are compared with every other model and combined with the Balanced Integration Score (Liesefeld & Janczyk, 2019); 50 is average. Single step shows how much reasoning a model needs to understand and solve one problem, multi-step how efficiently it keeps working on a task by building on its earlier reasoning, as in agentic workflows. Not part of StealthMark.

BENCHMARK RESULTS

Compare results

This link opens the same results in the same order.

StealthMark benchmark results

Reset benchmark preferences?

Return to the default view, clear comparison selections and restore the advertisement.