Reka Core is a powerful frontier-class multimodal language model approaching the capabilities of models like GPT-4 and Claude 3 Opus. It excels at both language and vision tasks, performing competitively on benchmarks like MMLU, GSM8K, and MMMU. Core outperforms many larger models on multilingual and long-context tasks. In blind human evaluations, it ranks highly for both text-only and multimodal chat. Core is well-suited for developers needing state-of-the-art performance on a wide range of language and vision tasks, especially those requiring strong multilingual capabilities or reasoning over long contexts. Its multimodal abilities make it a good fit for applications involving image and video understanding. While not always surpassing top models like GPT-4, Core offers very competitive performance and may be a strong choice for those seeking an alternative to the most prominent AI companies' offerings.