# Inkling 952b and Inkling Small 276b

**URL:** <https://forums.developer.nvidia.com/t/inkling-952b-and-inkling-small-276b/377238>\
**Category:** DGX Spark / GB10\
**Tags:** llm\
**Created:** [July 17, 2026, 11:23am UTC](https://forums.developer.nvidia.com/t/inkling-952b-and-inkling-small-276b/377238 "2026-07-17T11:23:23Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![eh17](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@eh17](https://forums.developer.nvidia.com/u/eh17)\
**Post date:** [July 17, 2026, 11:23am UTC](https://forums.developer.nvidia.com/t/inkling-952b-and-inkling-small-276b/377238/1 "2026-07-17T11:23:23Z")

</div>

Hello fellow Sparkists :)

The former CTO of OpenAI presented a very interesting new model today named Inkling.

The Inkling 975b is already online and i hope the Inkling Small 276b will be follow soon.

I guess the 276b Model in NVFP4 should run perfectly on Dual Spark :)

I am looking forward to test.

> **[thinkingmachines/Inkling · Hugging Face](https://huggingface.co/thinkingmachines/Inkling)**
>
> We’re on a journey to advance and democratize artificial intelligence through open source and open science.

> **[Inkling: Our open-weights model](https://thinkingmachines.ai/news/introducing-inkling/)**
>
> Our first open-weights model: multimodal, Mixture-of-Experts, with controllable reasoning effort. Available to fine-tune on Tinker.

Our model, called Inkling, is a Mixture-of-Experts transformer with 975B total parameters, 41B active. It supports a context window of up to 1M tokens. It was pretrained on 45 trillion tokens of text, images, audio and video. It is the first in a family of models of different sizes: alongside it we are sharing a preview of Inkling-Small, a lighter-weight model with 12B active parameters, trained with a similar recipe, that achieves strong performance with even lower cost and latency.

Inkling reasons natively over text, images, and audio, and balances cost with performance through efficient and controllable thinking effort. We trained it to be a broad, balanced foundation model: strong across many domains, flexible enough to adapt. Inkling is not the strongest overall model available today, open or closed. Instead, a combination of qualities makes it a good open-weights base for customization: multimodal capabilities, efficient thinking, and availability on Tinker for fine-tuning. Inkling is just the start: our first release in a model family we will continue to build on.

> **[Inkling-Small - Inkling: Our open-weights model](https://thinkingmachines.ai/news/introducing-inkling/#inkling-small)**
>
> Our mission is to build AI that extends human will and judgment. We have developed a platform that lets anyone customize models, previewed an AI system built for interactive collaboration, and published novel research. Today we are advancing our...

## **Inkling-Small** [#](https://thinkingmachines.ai/news/introducing-inkling/#inkling-small "Link to this section")

Alongside Inkling we are sharing a preview of Inkling-Small, a 276B-parameter Mixture-of-Experts model (12B active, vs. 41B for Inkling) with a different performance/latency trade-off. Inkling-Small matches or exceeds its larger sibling on many benchmarks — the result of improvements we made to the pre-training data and recipe for the smaller model. The two models share the same scalable post-training stack applied on top.

Best Regards

Erik

---

<div class="post-metadata">

**Author:** ![greg190](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@greg190](https://forums.developer.nvidia.com/u/greg190)\
**Post date:** [July 17, 2026, 3:46pm UTC](https://forums.developer.nvidia.com/t/inkling-952b-and-inkling-small-276b/377238/2 "2026-07-17T15:46:48Z")

</div>

we are working on standing this up on a cluster of 8 sparks