Sakana AI Open Models: Model List and Licenses
About Sakana AI
Sakana AI is a Japanese AI company. What it publishes on its SakanaAI organization on Hugging Face is mainly research models that test new training methods and architectures, rather than large production models.
What It Releases
- TinySwallow-1.5B (January 2025): a small Japanese model made with TAID (Temporally Adaptive Interpolated Distillation), its own knowledge distillation method. According to the model card, it was distilled with Qwen2.5-32B-Instruct as the teacher and Qwen2.5-1.5B-Instruct as the student, and further tuned for Japanese instruction following.
- RLT (June 2025): 7B and 32B student models trained with the Reinforcement-Learned Teachers pipeline, which trains teacher models with reinforcement learning.
- SparseLM (March 2026): research language models from 0.5B to 2B (no model card yet).
- Math (April 2026): a model based on gpt-oss-120b, fine-tuned for the AI Mathematical Olympiad (AIMO3).
- Others include research models for Continuous Thought Machines (CTM) and Llama-3-Karamaru, a conversational model that responds in the style of Edo-period Japanese (continually pre-trained on more than 25 million characters of Edo-period text).
Licenses
Most are Apache-2.0. Models based on Llama 3 (Karamaru, CycleQD and others) use the original Llama 3 license, and TAID-VLM is MIT. Read each model’s license text before use. This summary is not legal advice.
Running Them Locally
- The only official GGUF is TinySwallow-1.5B-Instruct (as of 2026-09-26). An official MLC build for running it in the browser is also available.
- Most models are research models and are not necessarily suited to general chat. For trying Japanese conversation locally, TinySwallow is the easy option.
Sources: the model card of SakanaAI/TinySwallow-1.5B-Instruct; the model list of the SakanaAI organization on Hugging Face (all as of 2026-09-26).
Our Coverage and Data
Local Model Watch has not published an article on Sakana AI’s models yet. The model list below is collected by code from Hugging Face every day. Part of our publisher index; model families link to their family pages.
At a Glance
| Item | Value |
|---|---|
| Hugging Face organization | SakanaAI |
| Models on Hugging Face | 49 |
| Articles on its own models | 0 |
| Articles on derived models | 0 |
Recent Models on Hugging Face
The publisher lists 49 models in total on Hugging Face. Below are up to 15 of the most recent generative models (text, image, video, audio and so on), the scope of this site, per organization. Classifiers, feature extractors, research components and repositories without a model card are omitted, and the publisher’s own quantized builds and helper variants are folded into the row of the original model. Parameter counts are left blank for repositories that pack weights into integer types, where Hugging Face’s count is not the real parameter count.
→ Scroll horizontally to see all columns
| Released | Model | Task | Parameters | License | Official variants |
|---|---|---|---|---|---|
| 2026-04-27 | SakanaAI/gpt-oss-120b-sft-aimo3-fishmath | text generation | — | apache-2.0 |
— |
| 2026-04-01 | SakanaAI/RePo-OLMo2-7B-stage2-L10 | text generation | 7.4B | apache-2.0 |
— |
| 2026-03-24 | SakanaAI/SparseLM2B | — | 2.2B | apache-2.0 |
— |
| 2026-03-24 | SakanaAI/SparseLM1B | — | 1.1B | apache-2.0 |
— |
| 2026-03-24 | SakanaAI/SparseLM1.5B | — | 1.6B | apache-2.0 |
— |
| 2026-03-24 | SakanaAI/SparseLM0.5B | — | 612M | apache-2.0 |
— |
| 2025-12-14 | SakanaAI/RePo-OLMo2-1B-stage2-L5 | text generation | 1.5B | apache-2.0 |
— |
| 2025-10-13 | SakanaAI/ctm-maze-large | text generation | 32M | apache-2.0 |
— |
| 2025-10-13 | SakanaAI/ctm-imagenet | text generation | 186M | apache-2.0 |
— |
| 2025-06-22 | SakanaAI/RLT-32B | text generation | 32.8B | apache-2.0 |
— |
| 2025-06-21 | SakanaAI/RLT-7B | text generation | 7.6B | apache-2.0 |
— |
| 2025-03-19 | SakanaAI/Llama-3-Karamaru-v1 | text generation | 8.0B | llama3 |
— |
| 2024-12-26 | SakanaAI/TinySwallow-1.5B | text generation | 1.5B | apache-2.0 |
Instruct-GGUF, Instruct-q4f32_1-MLC, Instruct |
| 2024-11-20 | SakanaAI/Metom | — | 6M | apache-2.0 |
— |
| 2024-08-15 | SakanaAI/TAID-VLM-2B | vision-language | 2.2B | mit |
— |
From the Hugging Face API. Last changed 2026-09-27 (JST). Collected by code.
Last updated 2026-09-27 (JST). The explanation at the top of this page was written with the help of AI from the primary sources it cites. The tables and lists under “Our Coverage and Data" are assembled by code from our article log and the Hugging Face API.