en.wikipedia.org/wiki/Inkling_(large_language_model)
1 correction found
It reasons natively from input in any of these four formats
This overstates Inkling’s supported input modalities. Thinking Machines says Inkling accepts text, image, and audio inputs; video appears in the training data, but is not listed as a supported input format.
Full reasoning
Thinking Machines’ own documentation distinguishes training data types from supported input modalities.
- The official model card says the training data "includes ... text, images, audio, video," but separately lists Inkling’s input modalities as only text, image, and audio.
- The official launch post likewise says: "Inkling reasons natively over text, images, and audio" — not video.
So the article’s wording is incorrect because it implies Inkling can directly take input in all four formats named in the previous sentence, including video. According to the vendor’s documentation, video was part of training data, but it is not one of the model’s supported input modalities.
2 sources
- Inkling Model Card - Thinking Machines Lab
Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs... Input modalities: Text... Image... Audio... Training data includes a broad variety of content types, including text, images, audio, video.
- Inkling: Our Open-Weights Model - Thinking Machines Lab
It was pretrained on 45 trillion tokens of text, images, audio and video... Inkling reasons natively over text, images, and audio.