Mercury 2.5 LLM hits 770 tokens per second (artificialanalysis.ai)

Mercury 2.5, a proprietary language model released by Inception in September 2026, is making waves in the AI/ML community by achieving a remarkable output rate of 770 tokens per second. As a reasoning model, it demonstrates above-average intelligence, scoring 12 on the Artificial Analysis Intelligence Index, which places it at the forefront of its category. With a context window of 260,000 tokens, it allows for extensive processing of text inputs, making it suitable for complex tasks across various fields, including coding and professional document reasoning. The model is significant not only because of its speed and efficiency but also due to its competitive pricing, costing $0.25 per million input tokens and $0.75 per million output tokens. These rates are on par with industry standards but reflect improved value when considering its high performance. Additionally, the model's ability to handle extensive context may enhance its utility in Retrieval Augmented Generation (RAG) workflows, providing a robust tool for developers and enterprises looking to leverage advanced AI capabilities in dynamic environments.