Thinky выпустила открытую модель Inkling 975B-A41B по лицензии Apache 2.0

Thinky released Inkling in July 2026 — a multimodal language model with 975 billion parameters and open weights under the Apache 2.0 license. This is the company's first full-scale large language model release, and it arrived with open weights right away.

What kind of model is Inkling

Inkling is built on a Mixture-of-Experts architecture: of the 975 billion parameters, only 41 billion are active on each request (hence the designation 975B-A41B). An MoE model routes each token to a subset of "experts," so the size is flagship-level while inference costs resemble those of a much smaller dense model. For users, this means Inkling responds like a system with 975 billion parameters, but in speed and cost per request it's closer to a 41-billion-parameter model — this is the balance that the largest labs use MoE to achieve. Inkling is multimodal — it works with more than just text.

Along with the flagship, Thinky also released a lighter version — Inkling-Small, with 276 billion parameters and 12 billion active (276B-A12B). Both models use the same MoE approach and differ in size; the smaller one is aimed at those without the hardware for the larger one.

• Flagship: Inkling, 975B total / 41B active (975B-A41B), multimodal
• Compact version: Inkling-Small, 276B / 12B active (276B-A12B)
• License: Apache 2.0 — open weights, commercial use without restrictions
• Positioning: presented as the best American open model under Apache 2.0
• Developer: Thinky, for which this is the first full-scale LLM release

Why open weights matter

Open weights mean anyone can download the model, run it on their own hardware, fine-tune it, and build it into products without going through a vendor's API. The Apache 2.0 license is one of the most permissive: it allows commercial use, modification, and distribution with virtually no restrictions, unlike the stricter "community" licenses used by some competitors. For startups and corporations, this removes the legal questions that keep some open models out of production.

According to Latent Space's AI News newsletter, the release turned out strong.

"Thinky's first full LLM release is a banger, and it comes with open weights as a bonus,"

Latent Space's AI News newsletter says.

Its place among open models

Inkling is positioned as the best American open model under the Apache 2.0 license — that's how the source describes it. Over the past two years, the bar for open weights has mostly been set by Chinese labs: DeepSeek, Qwen, Kimi. A strong American model with 975 billion parameters and a fully open Apache license marks a notable shift in that lineup.

What matters here is not just the fact of open weights, but the license itself. Many popular open models are released under their own "community" licenses with restrictions on the number of users or use cases. Apache 2.0 contains no such caveats, so an American model released specifically under this license removes some of the legal risk for businesses.

The pair of larger and smaller models (975B-A41B and 276B-A12B) covers different scenarios: the flagship for maximum quality, Inkling-Small for running on more modest hardware.

What this means

An open multimodal model of this size from an American team gives developers another strong option that can be deployed locally and used commercially without licensing caveats. Having two sizes available at once lets developers choose a model to fit their compute budget, and for the open-weight market it signals rising competition beyond Chinese labs.

Frequently asked questions

What is Inkling 975B-A41B?

It's Thinky's flagship multimodal model built on a Mixture-of-Experts architecture: 975 billion parameters total and 41 billion active per request. The weights are open under the Apache 2.0 license.

Is Inkling free?

Yes, the weights are released openly under Apache 2.0 — the model can be downloaded, run, and used in commercial products for free. You only pay for your own compute hardware or cloud.

How does Inkling-Small differ from the larger model?

Inkling-Small is smaller: 276 billion parameters versus 975 billion, and 12 billion active versus 41 billion. It's designed to run on less powerful hardware, while the larger model is built for maximum quality.