Inkling Small

Open-weight multimodal mixture-of-experts model with 12B active parameters from 276B total and 524K token context.

Strengths

Efficient inference from a sparse parameter architecture while maintaining multimodal capabilities across long contexts.

Best for

Applications where reduced computational requirements and faster latency are prioritized over the scale of the full Inkling model.

Limitations

Smaller active parameter count compared to Inkling means reduced reasoning depth and knowledge capacity for complex tasks.