Modelpublic

Zhongzhi1228/Qwen3.5-27B-RL

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face

1likes14downloads

Model Card

Qwen3.5-27B-RL

Qwen3.5-27B-RL is a reinforcement-learning checkpoint derived from Qwen/Qwen3.5-27B and associated with the Recursive Task Synthesis project.

Model details

—Architecture: Qwen3.5 causal language model with vision encoder
—Parameters: 27.8B
—Weight dtype: BF16
—Context length: 262,144 tokens
—Format: Hugging Face Transformers, sharded Safetensors

Related resources

Usage

python

from transformers import AutoModelForMultimodalLM, AutoProcessor

model_id = "Zhongzhi1228/Qwen3.5-27B-RL"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForMultimodalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype="auto",
)

License

The fine-tuned checkpoint weights and original repository contributions are released under the Creative Commons Attribution 4.0 International License. See LICENSE for the complete legal text.

This checkpoint is derived from Qwen/Qwen3.5-27B, which is licensed under Apache License 2.0. The unmodified base-model license is preserved in LICENSE.base-model and continues to apply to base-model components. The CC BY 4.0 grant does not supersede applicable base-model or other third-party terms.

The associated training-data repositories document their own upstream attribution and modification notices.

This license covers the original contributions and adaptations that the publisher has authority to license. It does not grant rights over third-party materials beyond what is permitted by their original terms. No endorsement by an upstream author or organization is implied.