CyberAgent Open Models: Model List and Licenses
About CyberAgent
CyberAgent is a Japanese internet company. It has released Japanese language models on its cyberagent organization on Hugging Face since 2023.
It releases both models trained from scratch in-house (CALM) and models built by further training other companies’ models for Japanese.
What It Releases
- CALM: Japanese language models trained in-house, from open-calm (May 2023) through calm2-7b (November 2023) to calm3-22b-chat (July 2024). According to its model card, calm3 was pre-trained from scratch on 2.0 trillion tokens.
- Japanese versions of other models: Llama-3.1-70B, Mistral-Nemo and DeepSeek-R1-Distill-Qwen (14B and 32B), further trained for Japanese.
- The CAT series (2026): CAT-Thinking-8B (May) is trained by reinforcement learning to think in Japanese (its reasoning trace is in Japanese) and is based on Qwen3-Swallow v0.2 from the Institute of Science Tokyo. CAT-Paws-8B (June) is an agent model on the same base, for multi-turn tasks involving coding, terminal operation and tool use. There is also CAT-Translate (0.8B, 1.4B and 3.3B) for translation.
- It also has research models for image generation and design (type-r, markupdm and others).
Licenses
They differ by model. CAT-Thinking, CAT-Paws, calm3 and calm2 are Apache-2.0; CAT-Translate and the Japanese DeepSeek-R1-Distill models are MIT; the Japanese Llama-3.1-70B uses the original Llama 3.1 license; and the early open-calm models are CC BY-SA 4.0. Read each model’s license text before use. This summary is not legal advice.
Running Them Locally
- There are no official GGUF builds (as of 2026-09-26); look for third-party conversions.
- The CAT models are 8B, an easy size for consumer GPUs. The CAT-Thinking model card recommends setting the maximum output tokens to at least 4096, since it thinks longer on difficult problems.
Sources: model cards of cyberagent/CAT-Thinking-8B and cyberagent/CAT-Paws-8B; the model list of the cyberagent organization on Hugging Face (all as of 2026-09-26).
Our Coverage and Data
Local Model Watch has not published an article on CyberAgent’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 | cyberagent |
| Models on Hugging Face | 32 |
| Articles on its own models | 0 |
| Articles on derived models | 0 |
Recent Models on Hugging Face
The publisher lists 32 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 |
|---|---|---|---|---|---|
| 2026-06-23 | cyberagent/CAT-Paws-8B | text generation | 8.2B | apache-2.0 |
— |
| 2026-05-28 | cyberagent/CAT-Thinking-8B | text generation | 8.2B | apache-2.0 |
— |
| 2026-02-12 | cyberagent/CAT-Translate-3.3b-beta | — | 3.4B | mit |
— |
| 2026-02-02 | cyberagent/CAT-Translate-1.4b | — | 1.4B | mit |
— |
| 2026-01-05 | cyberagent/markupdm | text generation | 7.5B | apache-2.0 |
— |
| 2025-12-09 | cyberagent/CAT-Translate-0.8b | — | 793M | mit |
— |
| 2025-04-23 | cyberagent/type-r | image generation | — | apache-2.0 |
— |
| 2025-02-10 | cyberagent/calm3-22b-chat-selfimprove-experimental | — | 22.5B | apache-2.0 |
— |
| 2025-01-27 | cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese | text generation | 32.8B | mit |
— |
| 2025-01-27 | cyberagent/DeepSeek-R1-Distill-Qwen-14B-Japanese | text generation | 14.8B | mit |
— |
| 2024-08-30 | cyberagent/Mistral-Nemo-Japanese-Instruct-2408 | text generation | 12.2B | apache-2.0 |
— |
| 2024-07-26 | cyberagent/Llama-3.1-70B-Japanese-Instruct-2407 | text generation | 70.6B | llama3.1 |
— |
| 2024-07-01 | cyberagent/calm3-22b-chat | text generation | 22.5B | apache-2.0 |
— |
| 2024-04-16 | cyberagent/opencole-stable-diffusion-xl-base-1.0-finetune | image generation | 2.6B | openrail++ |
— |
| 2023-11-01 | cyberagent/calm2-7b | text generation | — | apache-2.0 |
chat-dpo-experimental, chat |
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.