Taby's local brain
Taby uses fine-tuned Gemma-family models in three sizes. Smaller models keep everyday local help quick, while Taby 12B gives stronger answers when more memory is available
Updated July 27, 2026
The lineup
Each model powers the same local-first Taby experience for tasks, notes, planning, and assistant replies. The difference is how much speed, capability, and computer memory you want to balance
2B
Model 1 of 3
Taby 2B is the lightest fine-tuned Gemma-family model in the Taby lineup. It is designed for quick assistant replies, simple tasks, reminders, and short notes on lighter hardware
4B
Model 2 of 3
Taby 4B is the balanced everyday model. It keeps local replies quick while giving Taby more room to help with tasks, notes, planning, summaries, and questions about your day
12B
Model 3 of 3
Taby 12B is the strongest local model in the lineup. It gives the assistant more capacity for richer answers and planning help, but it needs more memory than the smaller models
The B means billions of learned parameters, approximately. A larger parameter count gives a model more capacity, but it also asks more from your computer
Model size is not the whole answer. Fine-tuning, quantization, context length, available memory, and the local AI backend all affect how fast and useful a model feels
Taby's models are built to run on your computer, so everyday assistant work can stay local. Optional cloud AI is available for harder jobs when you choose to use it
These are language models. Voice input and spoken replies also depend on Taby's separate speech-to-text and text-to-speech components
Hardware fit
Model size is only part of the answer. Available memory, processor, graphics, operating system, and local speed all affect how a model feels. Run the browser test for a practical estimate, then correct anything the browser could not detect
Test this computerQuick answers
No, these are different local model sizes for the same Taby assistant experience
Taby uses fine-tuned models from the Gemma family in 2B, 4B, and 12B sizes
No, smaller models can keep replies quicker on lighter hardware, while Taby 12B can give stronger answers when enough memory is available
They are built to run locally on your computer, while cloud AI remains optional for harder jobs when you choose it
They generate Taby’s language responses, while listening and spoken audio use separate speech-to-text and text-to-speech components
Use the local AI compatibility checker to estimate a sensible starting tier for your computer, then correct any hardware detail the browser could not detect
What's next
Use one ready-to-go local assistant, or choose a model file and build the setup yourself
Ready to use
Taby is the all-in-one local option: the model, chat, tasks, notes, planning, and voice work together in one assistant on your computer
See what Taby can doDIY setup
Choose a GGUF model file on Hugging Face, then run it locally with Ollama, LM Studio, or llama.cpp. You manage the runtime and updates
Browse GGUF models