Reka Edge

About

Reka Edge is a 7-billion parameter multimodal language model created by Reka AI, designed for deployment on a variety of devices, including laptops and high-end tablets. It skillfully supports text, image, video, and audio inputs. Although its overall performance is considered below average, scoring 0.657 on the MMLU benchmark, it surpasses other models of similar size like Mistral-7B and Gemma-7B on specific benchmarks such as MMLU, GSM8K, and HumanEval. The model features a vast 64k context window and delivers fast processing speeds at 35.3 tokens per second with low latency. It is available via API, priced affordably at $0.10 per million tokens. Reka Edge was trained on approximately 4.5 trillion deduplicated tokens, utilizing diverse data sources, and features a knowledge cutoff of November 2023.

Capabilities

Providers(1)

API Versions