Welcome Inkling 2 Mini, Cutting-edge Performance for your Local Setup
Blog post from Hugging Face
Inkling 2 Mini is a multimodal mixture-of-experts model designed for local use, with about 26 billion total parameters and 3 billion active per token, support for text and images, and a context window of up to one million tokens. Its 24 physical decoder layers include a 12-layer core that runs three times, reusing weights while passing information between iterations; a separate research version allows fewer passes for faster, potentially lower-quality generation. The post reports strong results across several benchmark categories, including coding, tool use, vision, reasoning, and safety, while noting that further validation is needed. It describes support for Transformers, llama.cpp, MLX, and SGLang, along with mixed-precision weights, 4-bit and 8-bit quantized versions, and optional speculative decoding. Demonstrations include procedural graphics, an animated recreation of Van Gogh’s “The Starry Night,” a trip-expense app built from a sketch, and a plant-therapy website, illustrating the model’s use for coding and creative tasks.
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.