A large open base for researchers who plan to fine-tune; its makers admit it isn't the strongest model available.

Kenning's take

What it does

Inkling is a 975 billion parameter mixture-of-experts model from Thinking Machines Lab, with 41 billion parameters active and up to a 1 million token context. It reasons over text, images and audio, and its thinking effort can be turned up or down to trade speed for quality. The lab says it is not the strongest model available but is meant as an open base for customisation, with weights on Hugging Face under Apache 2.0 and fine-tuning on the Tinker platform.

• 975B total parameters, 41B active
• Up to 1M token context (64K or 256K on Tinker)
• Text, image and audio input
• Controllable thinking effort
• Fine-tuning and playground on Tinker

Who it suits. Developers and researchers who want a large open multimodal model to fine-tune for their own use.

Key facts

Category

AI models

Made by

Thinking Machines Lab

Released

15 July 2026

How to get it

open weights on Hugging Face (Apache 2.0); fine-tuning, sampling and a playground on Tinker

Works on

Hugging Face, Tinker, self-hosted

Free plan

Weights free to download under Apache 2.0

API price

Tinker: $3.74 per million prefill tokens, $9.36 per million sampled, $11.22 per million trained (list prices; limited-time 50% discount)

Models and features

• Inkling-Small 2026-07-30

FAQ

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.