Reka Core

Reka Core is the largest and most capable model in Reka AI's multimodal family, unveiled in April 2024 by a team of former Google DeepMind researchers. It processes text, images, video, and audio within a single transformer architecture with up to 128K tokens of context, and was positioned at launch as one of only two commercially available comprehensive multimodal solutions.

At release, Reka reported Core performing at par with GPT-4, Claude 3, and Gemini models across MMLU, math, coding, reasoning, and image understanding benchmarks, with particular strength in video question answering where Core and Flash were reported to outperform Gemini Ultra. Reka AI later built its Nexus enterprise AI workforce platform on top of the Reka model family.