Core Design-Multimodal language model

Reka AI launched its largest and most powerful multimodal language model, Reka Core. Core is capable of processing text, image, video, and audio inputs.

Reka AI provides world-class multimodal language models, available in three sizes, that support a range of needs. These models can process image, video with audio, text, and documents, and offer leading reasoning and coding capabilities. They are highly flexible in deployment and have a 128K context window, with multilingual support. The models are used in various industries, including commerce, financial services, healthcare, social media, digital content, and robotics.

Reka is available in three different sizes – Reka Core, Flash and Edge.

⏺ Edge: 7B lightweight/local model

⏺ Flash: 21B, fast speed, powerful functions

⏺ Core: The largest model, capable of handling complex tasks

During the pre-training phase, the model processed 32 languages ​​including Chinese, Japanese, French, Korean and Spanish, and has strong cross-language processing capabilities.

You can log in with your Google account and try it out in Playground.

Link: click to open