TinySwallow-1.5B-Instruct
Sakana and the Swallow team add Japanese instruction tuning to the Qwen-derived TinySwallow model. The release includes synthetic dialogue data associated with Gemma.
Research release
Read both the license section and intended-use statement
The license section permits commercial use when both sets of terms are met. A separate use section calls the model an experimental prototype and says commercial deployment is not intended. The catalog preserves both statements.TinySwallow-1.5B-Instruct official model card
Working with this release
Practical guidance. Apply the supplied chat template. Assess Japanese conversation quality separately from the memory savings offered by a particular weight format.
GGUF and MLC are packaging choices
Sakana publishes GGUF weights and a q4f32_1 MLC package for this instruction model. They are grouped here as formats rather than separate trained models.TinySwallow-1.5B-Instruct-GGUF official model cardTinySwallow-1.5B-Instruct-q4f32_1-MLC official model card
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | SakanaAI/TinySwallow-1.5B-InstructTinySwallow-1.5B-Instruct-GGUFTinySwallow-1.5B-Instruct-q4f32_1-MLC |
| Base | TinySwallow-1.5B; originally Qwen2.5-1.5B-Instruct.TinySwallow-1.5B-Instruct official model card |
| Teacher | Qwen2.5-32B-InstructTinySwallow-1.5B-Instruct official model card |
| Terms | The card names Qwen's Apache 2.0 terms and Gemma data terms and policies.TinySwallow-1.5B-Instruct official model card |
Related models
- TinySwallow-1.5B
Sakana AI ยท TinySwallow
A compact Japanese model co-developed with the Swallow team.
Launch date not confirmedResearch releaseDownloadable