Multimodal AI you can deploy anywhere

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

Try it out yourself.

Would you rather use our API? We’ve got you covered!

Our World-class Models - Core, Flash, Edge & Spark

Our models are efficiently trained from scratch across a vast fleet of thousands of GPUs. Each model in our lineup is natively multimodal, meticulously designed to meet a distinct set of requirements. Reka Core, our flagship model, rivals industry leaders such as OpenAI, Anthropic, and Google, demonstrating competitive performance across widely accepted evaluation metrics. On the other end of the spectrum, Edge and Spark are compact models optimized for on-device execution, while Flash, our Turbo-class offering, is pushed to its absolute limits to provide significant performance at fast speeds.

Learn more

Read our Technical report on Reka Core

Natively Multimodal

Reka’s multimodal models can perform many tasks such as generating textual descriptions and structured taxonomies from videos and images, translating speech and audio from and to many languages, and answering complex questions from long multimodal documents, writing and executing code, and many more.

Describe the setting, action, and character in this video

The video is set in a location that has a house with a pool overlooking the sea. The video shows someone jumping into an infinity pool from the edge of the house. They then swim in the pool. Regarding the person, they appear to be a woman. She is wearing a black swimsuit.

**Where was this taken?**
This image was taken in San Sebastián, also known as Donostia, a coastal city in the Basque Country, Spain. The perspective is from the promenade along the Bay of La Concha, looking towards the city center. The bay is one of the most iconic features of San Sebastián, and the cityscape with its historic buildings is a well-known view of the area. The presence of the sea wall and the style of the buildings, along with the mountainous backdrop, are characteristic of this region.

What is happening here?
The video shows a person using a table saw to cut a piece of wood. The person is wearing blue gloves and is using a push stick to guide the wood through the saw. The saw blade is visible and the wood is being held in place by clamps. The person adjusts the clamps and the wood before starting the saw.

Model capabilities

Each of our models can be fine-tuned and customized, including using your own proprietary data, securely for deployment in the cloud, on-premises, or on-device.

Image and video comprehension

Achieve a deeper understanding of your multimodal data and solve problems beyond text with our natively multimodal models
General reasoning

Understand and apply everyday knowledge to solve problems and tasks that require logical deduction and inference, making it suitable for complex problem solving
Multilingual understanding

End to end support for 32 languages in both text and spoken form
Tool use, function calling, and coding

Execute tasks, unlock agentic flows and automate AI driven solutions with our models’ function calling capabilities.
Speech input and output

Bring new products to live with the ability to comprehend and respond speech.

Advanced writing and reasoning capabilities

Our models have advanced reasoning capabilities to solve a wide range of complex problems.

Try Now

Imagine a plot for a crossover between Dune and Dora the Explorer
Dora and Boots stumble upon a mystical portal leading to Arrakis, where they join forces with Paul Atreides and the Fremen to find the hidden spice mines and thwart the Harkonnens' plan to dominate the universe, all while teaching valuable lessons in teamwork and problem-solving.
How can you stack a bowl, egg, ball and book in a stable way?
Begin with a flat book as the base. Place an upside-down bowl on the book, securing its edges. Nestle an egg inside the bowl against its curve to prevent rolling. Finally, center a ball on top of the bowl, distributing its weight evenly. This stack leverages each item's shape for stability.
What is the least number of coins required to make a total of USD 1.35?
The least number of coins required to make $1.35 USD using standard U.S. coin denominations (penny, nickel, dime, quarter) is 6 coins: 5 quarters and 1 dime.

Deployment Methods

With flexible deployment options centered around your privacy and cost-efficiency

API

All models are available via our API. Get started today and build the AI features you need with our models.
On-premise

Deploy our model weights on your own infrastructure. Maintain your data and privacy requirements.
On-device

Add low latency cutting edge multimodal intelligence to your devices with locally served model.

Explore Our Models

Try Now

Explore all our models on Reka chat. Try our image and video comprehension or solve complex problems.

Try Reka
Use our API

Start integrating our models into your products today with our API. No complex pricing plans, just pay for what you use.

Use our API