Enterprise AI player Cohere is making a loud entrance into the “AI for everyone” race with its new Tiny Aya family of multilingual models, a move clearly aimed at grabbing headlines and market share in one shot.

Branded as open-weight and “for the world,” Tiny Aya is being sold as the model that can run on everyday laptops while speaking more than 70 languages. The pitch is simple and flashy: powerful generative AI that doesn’t need the cloud, doesn’t need constant internet, and supposedly respects your privacy by staying on your device.

The company is leaning hard into the inclusion narrative. Tiny Aya is being framed as a direct answer to the dominance of English-first AI, with Cohere spotlighting South Asian languages like Bengali, Hindi, Punjabi, Urdu, Gujarati, Tamil, Telugu, and Marathi. The message: Big Tech ignored these users, and Cohere is here to fix it.

Under the hood, the headline number is a 3.35-billion-parameter base model. In an era obsessed with ever-bigger models, Cohere is spinning this relatively compact size as a “sweet spot” between performance and efficiency. On top of that, it’s rolling out TinyAya-Global, a version fine-tuned to follow user instructions across many languages, positioned as the general-purpose workhorse for everyday tasks.

But the real marketing hook is the regional branding. Cohere has sliced Tiny Aya into themed variants: TinyAya-Earth for African languages, TinyAya-Fire for South Asian languages, and TinyAya-Water for Asia Pacific, West Asia, and European languages. The company claims this regional focus delivers deeper cultural nuance and stronger linguistic grounding, while still covering a wide range of languages. The naming alone makes it clear Cohere wants these models to feel like a global franchise, not just another technical release.

Cohere is also making a point of how it trained these models. Instead of boasting about massive supercomputers, it highlights a single cluster of 64 Nvidia H100 GPUs. The narrative: you don’t need hyperscale infrastructure to build useful AI. That line is aimed squarely at researchers, startups, and enterprises that want capable models without paying frontier-model prices.

The offline angle is another key part of the story. By running locally, Tiny Aya is being positioned as the go-to solution for translation, privacy-preserving assistants, and specialized tools for communities that live and work primarily in their native languages. In regions with unreliable connectivity or high data costs, Cohere is clearly hoping Tiny Aya will be seen as the practical, on-the-ground alternative to cloud-only AI.

Distribution is also being used as a signal of seriousness. Tiny Aya is rolling out across major AI model hubs and Cohere’s own platform, bundled with training and evaluation datasets and an upcoming technical report. The company wants this to look like more than a product drop; it wants it to look like the start of a full ecosystem around multilingual, locally deployable AI.

In a crowded AI landscape dominated by a few giants, Tiny Aya is Cohere’s attempt to stand out with a mix of accessibility, regional branding, and multilingual ambition. Whether it’s a genuine shift in how AI reaches the world or just a well-packaged play for attention, one thing is clear: Cohere wants Tiny Aya to be seen as the model that finally takes AI beyond English and beyond the cloud.