Inkling Small (batch)

Open-weight multimodal mixture-of-experts model with 12B active parameters out of 276B total and a 524K token context window, optimized for batch processing.

Strengths

Routes computation efficiently across sparse parameters, reducing inference overhead while maintaining multimodal capabilities across text and images.

Best for

Batch processing jobs where latency is not a constraint and you want to process large volumes of multimodal content with lower compute requirements than the larger Inkling variant.

Limitations

The smaller active parameter count makes it less capable than Inkling on complex reasoning tasks; the batch-only interface means real-time or streaming use cases are not supported.