Multimodal AI you can deploy anywhere

Next-generation models to empower AI agents that can see, hear, and speak.

Natively Multimodal

Our models Reka Core (67B), Flash (21B), Edge (7B), and Spark (2B) are trained from scratch on text, code, images, video, and audio data with a novel multimodal architecture.

They are available in different sizes to power our products and meet various deployment settings, including on devices, on premises, and in the cloud.

See more details here or read our technical report to learn more about our innovative approach to train these models.

Extract person, location, key objects to a structured format

{
"person": "A young Asian man",
"location": "a brewery with metal kegs and stainless steel vats in the background",
"key_objects": [ "metal kegs on a wooden pallet", "stacks of boxes in the background" ]
}

Examples

Whom is this invoice billed to and how much is the amount?

This invoice is billed to Jackson Wang and the amount is HK$159.00.

Translate this audio file to Indonesian

Halo, apa kabar?

Our Partners

Join Our Team

Join us in this exciting journey and experience a new era of growth.