Deep Cogito: Cogito v2.1 671B

Deep Cogito: Cogito v2.1 671B MoE represents one of the strongest open models globally, matching the performance of frontier closed and open models. This advanced system is trained using self-play with reinforcement learning to achieve state-of-the-art performance across multiple categories, including instruction following, coding, longer queries, and creative writing. It demonstrates significant progress toward scalable superintelligence through policy improvement. This model offers a substantial 128K token context window, allowing for complex and extended interactions, and a max output of 4K tokens. It supports both code generation and streaming capabilities, making it highly versatile for various applications. Pricing is competitive at $1.25 per 1M input tokens and $1.25 per 1M output tokens, available through our PRO Access Tier.