Description
Inkling-Small is a general-purpose multimodal model designed to handle text, image, and audio inputs, producing text outputs. It is suitable for use in English and other languages, and supports multiple coding languages. Developers can utilize Inkling-Small to build AI-powered applications such as chatbots, coding assistants, and retrieval-augmented generation systems. The model is released with open weights, allowing for research, fine-tuning, and integration into third-party products.
Read More
The model architecture is a 42-layer decoder-only transformer with a sparse Mixture-of-Experts feed-forward backbone. Each token is routed to 6 of 256 experts, plus 2 shared experts active on every token. It features a hybrid attention mechanism combining local and global layers. Inkling-Small is natively multimodal, encoding images via a hierarchical patch encoder and audio via discrete token encoding, with all modalities projected into a shared hidden space.
Inkling-Small supports local deployment using open-source libraries such as SGLang, vLLM, TokenSpeed, Unsloth, and Huggingface. API access is available through third-party inference providers. The model's training data includes diverse content types sourced from publicly available repositories, third-party acquisitions, and synthetic generation.
Safety evaluations were conducted to ensure Inkling-Small's responsible use across text, audio, and image inputs. Mitigations were applied to reduce risks, and common moderation tools can be layered around the model to enhance safety. Limitations include potential biases and hallucinations, common to foundation models, and degraded performance in long conversations. Developers are encouraged to apply human oversight and additional safeguards for high-stakes applications.
- CategoryAI Models & LLMs
- Pricing ModelOpen source
- DeploymentCloud, Self-hosted
- PlatformsWeb
Inkling-Small Highlights
Multimodal model
Open weights
API access
Local deployment support
42-layer transformer
Sparse Mixture-of-Experts
Hierarchical patch encoder
Discrete token encoding
Getting Started with Inkling-Small
Access model page: visit Hugging Face
Authenticate: set up access credentials
Configure: adjust model settings
Send prompts: input text, image, or audio
Fine-tune: customize model outputs
Inkling-Small's Use Cases
- Chatbots
- Coding Assistants
- Multimodal Applications
- Instruction Following
- Research
