Reka AI Debuts Reka Core, a Leading Multimodal Language Model
Reka AI, a startup known for customizing AI models for large enterprises, has introduced Reka Core. This advanced multimodal language model is capable of processing various types of prompts, including text, images, videos, and audio.
Founded in 2022 by researchers from major tech giants such as Google’s DeepMind, Baidu, and Meta, Reka AI has already secured $60 million from several investors, including New York Life Ventures, Radical Ventures, Snowflake Ventures, DST Global, and entrepreneur Nat Friedman.
Reka Core is considered the most substantial model developed by the company. It showcases remarkable skills that could rival those of larger, well-funded models from competitors such as OpenAI, Anthropic, and Google. Specifically, the company claims that Reka Core performs exceptionally on benchmarks like the Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU), a comprehensive evaluation framework that processes and interprets information from various sources such as text, images, videos, and audio, surpassing GPT-4V and other models on various tasks.
Furthermore, the model is capable of understanding and processing video inputs, which positions it ahead of competitors like Gemini Ultra. It also displays strong abilities in reasoning and code generation, in addition to offering a chatbot interface for free.
However, it is worth noting that Reka Core has some limitations when it comes to handling long prompts. For instance, the free version of this model can only process up to 4000 tokens, compared to other models that offer standard context windows of up to 128,000 tokens.
Reka Core was trained from scratch using data from 32 languages, making it fluent and knowledgeable in many uncommon languages and dialects. However, it is not an open-source model. The company is collaborating with leading global technology platforms and government organizations such as Snowflake, Oracle, and AI Singapore.
In testing, Reka Core demonstrated notable visual analysis capabilities and strong language comprehension skills. However, its creative writing abilities were consistent but not exceptional compared to other AI models like Claude. It also incorporates built-in safety controls to prevent the generation of harmful or unethical content.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.