Sakana AITAID
TAID-LLM-1.5B
TAID-LLM applies Sakana's temporally adaptive distillation to Qwen2. The student starts from Qwen2-1.5B-Instruct and learns from Qwen2-72B-Instruct.
Research release
The teacher and student are different sizes
The card names Hugging Face's SmolLM corpus as training data and identifies English as the release language. This is a Sakana-trained derivative, not an unchanged Qwen mirror.TAID-LLM-1.5B official model card
Working with this release
Practical guidance. Use English task examples when reproducing the documented scope. Compare the adapted student with its original Qwen checkpoint before attributing a benefit to distillation.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | SakanaAI/TAID-LLM-1.5B |
| Student ancestry | Qwen2-1.5B-InstructTAID-LLM-1.5B official model card |
| Teacher | Qwen2-72B-InstructTAID-LLM-1.5B official model card |
| Terms | Apache 2.0; the card separately describes a research prototype not intended for commercial or critical deployment.TAID-LLM-1.5B official model card |
Related models
- TAID-VLM-2B
Sakana AI · TAID
A compact vision-language model distilled from InternVL2-8B.
Launch date not confirmedResearch releaseDownloadable - TinySwallow-1.5B
Sakana AI · TinySwallow
A compact Japanese model co-developed with the Swallow team.
Launch date not confirmedResearch releaseDownloadable