Reka Flash is a multimodal language model developed by Reka AI, designed to provide a balance between high reasoning performance and low-latency inference. Positioned as the mid-tier offering in the Reka model family—between the high-capability Reka Core and the compact Reka Edge—Flash is optimized for enterprise applications that require efficient processing of complex, multi-format data.
The September 2024 update to Reka Flash introduced significant refinements to its core capabilities, particularly in coding, mathematical reasoning, and document extraction. As a natively multimodal model, it can process and reason across text, images, and video content. It is capable of analyzing video files up to one minute in length and handling large visual documents with high precision.
Technical Capabilities
The model features a 128,000-token context window, enabling it to ingest and synthesize information from extensive datasets or long-form media. Its architecture is built to support high-throughput environments, making it suitable for real-time interactions and agentic workflows. While the model is primarily optimized for English, it demonstrates competitive performance in multilingual benchmarks and cross-lingual understanding.