AI21 Labs

Jamba 1.5 Large is a hybrid Transformer-Mamba model designed for exceptional long-context performance, supporting up to 256K tokens. It offers a strong balance of speed, quality, and cost-efficiency, and includes multimodal capabilities for processing both text and images.

Jamba 1.5 Mini is a compact, efficient model from AI21 Labs' Jamba series, utilizing a hybrid Transformer-Mamba architecture. It is optimized for fast inference and low cost, making it suitable for high-throughput applications requiring strong reasoning and code generation capabilities.

Jamba 1.6 Large is a large-scale hybrid model from AI21 Labs, combining Transformer and Mamba (SSM) architectures. It is optimized for extremely long context windows and high throughput, offering strong performance on complex reasoning and generation tasks.

Jamba 1.6 Mini is a compact and efficient hybrid model from AI21 Labs, combining Transformer and Mamba architectures. It is optimized for fast inference and supports very long context windows, making it suitable for processing large documents or complex conversations.

Jamba 1.7 Large is a large-scale hybrid model from AI21 Labs that integrates Transformer and Mamba architectures to achieve efficient long-context processing and robust reasoning capabilities.

Jamba 1.7 Mini is a lightweight, efficient model from AI21 Labs' Jamba series, optimized for fast inference and cost-effectiveness. It leverages a hybrid Transformer-Mamba architecture to deliver strong performance on long-context tasks while maintaining low latency.

Jamba Reasoning 3B is a compact reasoning model from AI21 Labs, optimized for logical and analytical tasks. It features a 3-billion parameter architecture designed for efficient performance on reasoning benchmarks.