Ornith AI releases 122B MoE multimodal model under MIT

Ornith AI has published Ornith-1.5-122B on Hugging Face, a 122-billion-parameter mixture-of-experts model that handles both text and images. The release lands under an MIT license, one of the most permissive terms available for a model of this scale.

According to the model card, Ornith-1.5 is built on the Qwen3.5 MoE architecture. That lineage matters: rather than training a frontier base from scratch, Ornith AI is adapting a well-regarded open foundation and extending it with multimodal capabilities, positioning the model as a general-purpose text-and-vision system.

Why it matters

Mixture-of-experts designs activate only a fraction of their total parameters for any given token, which lets a 122B-parameter model run more efficiently than a dense model of comparable size. Combined with an MIT license, that makes Ornith-1.5 an attractive option for teams that want commercial freedom without restrictive usage clauses.

A few things to keep in mind:

• The release covers both text (LLM) and vision-language (VLM) use cases, with text as the primary modality.
• Context length and detailed benchmark figures were not specified in the release record.
• The MIT license permits broad commercial and derivative use.

For developers evaluating open multimodal options, Ornith-1.5 is worth a look alongside the Qwen family it derives from. Full details and weights are available on the model's Hugging Face page.