Open source, open collaboration, open data governance, and inspectable technical paths. Model weights, evaluation methods, model cards, training notes, and core tools should be open to review.
About
Let AI speak Taiwan,hear Taiwan,and read Taiwan.
OpenFormosa is an open family of AI models. With the Gaji bag (Taiwan Striped Bag) as its symbol, it weaves Taiwan's voices, texts, images, and memories into model capabilities people can try, inspect, and deploy.
Why "OpenFormosa"?
The name is a working rule: keep the family of models open, inspectable, and grounded in Taiwan's language contexts.
Taiwan's language contexts include Traditional Chinese, Taiwan Mandarin, Taigi and Hakka accents, Bopomofo, Tâi-lô, local speech, document formats, and cultural memory.
Why the Gaji bag?
The Gaji bag (Taiwan Striped Bag) is an everyday object in Taiwan: cheap, durable, and instantly recognizable. It fits what OpenFormosa is trying to do: carry different voices, texts, images, and memories, then weave them into usable model capabilities.
It carries Taiwan corpora, voices, documents, knowledge, and use cases.
It lets model capability flow to researchers, developers, educators, partner organizations, and integrators.
Not a single demo, but a family of models that can be tuned, deployed, distilled, and extended.
A Taiwan symbol, but not a closed one. The models focus on Taiwan scenarios while staying open to inspection and use.
The weave mirrors the work itself: tokenizer design, data governance, evaluation, and model capability all need to connect.
Taiwan is not a translation patch.
Writing Traditional Chinese does not mean understanding usage, accents, and culture in Taiwan. Taiwan Mandarin, Taigi and Hakka accents, Bopomofo, Tâi-lô, local terms, and document formats need to be part of the model design.
Generic models often miss Taiwan names, institutions, mixed language sentences, local pronunciation, receipts, menus, public notices, and address formats.
We develop models, tokenizers, benchmarks, model cards, and release notes together, so Taiwan capability can be inspected rather than merely claimed.
Taiwan context is model work.
This is not about translating a generic model into Traditional Chinese. Taiwan terms, speech, documents, and cultural memory need to be designed into the tokenizer, training recipe, evaluation suite, and task adapters from the start.
Not just Chinese
A model can write Traditional Chinese and still miss Taiwan's institutions, place names, speech habits, internet tone, addresses, forms, and public language.
Small, but useful
A compact model matters when schools, labs, teams, and venues can run it, tune it, and deploy it privately.
Open builds trust
Open means people can inspect model cards, evaluation sets, tokenizer choices, training recipes, benchmark artifacts, and release notes instead of guessing what happened.
Shared base, separate tasks
ASR, TTS, and OCR should share Taiwan language knowledge, while each task keeps its own data, training path, and evaluation rules.
OpenFormosa is an open family of AI models. With the Gaji bag (Taiwan Striped Bag) as its symbol, it weaves Taiwan's voices, texts, images, and memories into model capabilities people can try, inspect, and deploy.
If you work with Taiwan language, speech, documents, or local use cases, join us on GitHub or in the community to discuss testing, docs, model integration, data licensing, and collaboration. If you only want to see the results, start with the demo and released models.
Make local usage, local accents, and mixed language contexts part of everyday model work.