Jamba 1.6 Large
Overview
Jamba 1.6 Large is a language model developed by AI21 Labs, released in March 2025. It utilizes a hybrid SSM-Transformer architecture, combining State Space Models (SSM) with Transformer attention mechanisms to enhance efficiency and performance. This model features a context window of 256,000 tokens, enabling it to process extensive text inputs effectively. With 94 billion active parameters and a total of 398 billion parameters, Jamba 1.6 Large is designed to handle complex tasks at an enterprise scale, supporting function calling, structured JSON output, and grounded generation. It is available under the Jamba Open Model License, allowing for both research and commercial use under specified terms. The model supports multiple languages, including English, Spanish, French, Portuguese, Italian, Dutch, German, Arabic, and Hebrew. Notably, Jamba 1.6 Large outperforms several competitors on quality benchmarks, demonstrating superior long-context handling, speed, and overall performance. Its open-weight nature and advanced capabilities make it a strong choice for applications requiring extensive context processing and efficient inference, positioning it favorably against peers in the market.