The world is complex. We help you understand it to better navigate it.

AI has the power to lower the time and effort required by organizations to make necessary decisions.
By lowering the need for human intervention, organizations become more productive, on a massive scale.

Reka multimodal AI models impact everyday businesses

Data is the lifeblood of the modern enterprise. Data is not only being generated and used in greater quantities, it is
increasingly unstructured and multimodal in nature. This data includes text, images, audio and even videos.
To make the best decisions and act on them, AI systems need to be able to understand fully and act precisely. Reka models help make sense of this data in ways previously not possible. Our models understand and analyze multimodal data to assist businesses in making decisions and taking actions.

1. Understand better

While text comprises a large amount of data within the enterprise today, there’s an increasing amount of multimodal
data. This is especially true in certain industries that generate and utilize this type of data more frequently. Multimodal
data is also not mutually exclusive from text. In fact, there are often components of text associated with images, videos
and audio. Reka models make it possible to analyze a wide range of multimodal input and synthesize organized outputs
with great accuracy and nuance.
Analyze and derive insights from unstructured / multimodal data

Understand multiple languages
We pre-trained our models in 32 languages. We focus on 12 major languages that address massive global population
demographics. These include English, Spanish, French, Italian, Arabic, Hindi, Korean, Japanese, Vietnamese, Tamil,
Bahasa Indonesia, and Thai. Our models understand input and can generate output in these languages in both text and
spoken form.

Recall and reason across ultra long context to for better and more intelligent insights

We support multimodal inputs with a large context window of up to 128K tokens (100K English words, 100 images and
1 hour video) with an option to extend it to 400K tokens (equivalent to ~300,000 English words). Our models have the
ability to reason over this long horizon. This enables applications that require understanding over multiple pieces of
information such as question answering over a large private dataset collection as well as provide a viable alternative to
recall and analyze information as well as take action.

2. Take action

Direct action from Reka models can be via text (of various formats) or speech. This allows us to have a dynamic range of natural language interactions and sequential actions. Express output via text, or speech

Run within your environment with license fee per server.

Note: Minimum of 2*A10 GPU or 2*A100 GPU are needed to run Reka Flash, please select the right GPU