Instructions to use yifanyu/DiffusionGemma-26B-A4B-it-ORL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yifanyu/DiffusionGemma-26B-A4B-it-ORL with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="yifanyu/DiffusionGemma-26B-A4B-it-ORL")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("yifanyu/DiffusionGemma-26B-A4B-it-ORL")
model = AutoModelForMultimodalLM.from_pretrained("yifanyu/DiffusionGemma-26B-A4B-it-ORL", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use yifanyu/DiffusionGemma-26B-A4B-it-ORL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yifanyu/DiffusionGemma-26B-A4B-it-ORL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yifanyu/DiffusionGemma-26B-A4B-it-ORL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/yifanyu/DiffusionGemma-26B-A4B-it-ORL

SGLang

How to use yifanyu/DiffusionGemma-26B-A4B-it-ORL with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "yifanyu/DiffusionGemma-26B-A4B-it-ORL" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yifanyu/DiffusionGemma-26B-A4B-it-ORL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "yifanyu/DiffusionGemma-26B-A4B-it-ORL" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yifanyu/DiffusionGemma-26B-A4B-it-ORL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use yifanyu/DiffusionGemma-26B-A4B-it-ORL with Docker Model Runner:
```
docker model run hf.co/yifanyu/DiffusionGemma-26B-A4B-it-ORL
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

DiffusionGemma-26B-A4B-it-ORL

DiffusionGemma + ORL: DiffusionGemma-26B-A4B post-trained with On-Policy Residual Learning (ORL); it decodes 3-6 more tokens per forward than DiffusionGemma at comparable quality

Highlights

Higher tokens per forward (TPF) on every benchmark: 30.0 vs 26.0 on GSM8K, 27.9 vs 24.5 on HumanEval, 28.7 vs 23.1 on MBPP
Comparable accuracy: equal or higher on GSM8K, MATH-500 and HumanEval, 0.3 points lower on MBPP
1,456 tokens/s for a single request on GSM8K with dlmrt on one H100; with NVFP4, 4,214 on one B200, 7,708 on eight B200s and 1,540 on a consumer RTX 5090

Results

Accuracy (%) / TPF; each model uses its selected decoder.

Model	GSM8K	MATH-500	HumanEval	MBPP
DiffusionGemma-26B-A4B	95.8 / 26.0	92.3 / 23.7	90.9 / 24.5	87.7 / 23.1
DiffusionGemma-26B-A4B-it-ORL	96.0 / 30.0	92.3 / 27.4	91.9 / 27.9	87.4 / 28.7

Speed

Tokens per second for a single request on GSM8K

Tokens per second for a single request, measured with dlmrt unless noted, with the paper's prompts and decoders (the paper reports the base model on GSM8K and HumanEval):

Model	GPU, precision	GSM8K	MATH-500	HumanEval	MBPP
DiffusionGemma-26B-A4B (vLLM)	H100, BF16	990	—	963	—
DiffusionGemma-26B-A4B	H100, BF16	1,271	—	1,196	—
DiffusionGemma-26B-A4B-it-ORL	H100, BF16	1,456	1,415	1,378	1,361
DiffusionGemma-26B-A4B-it-ORL-NVFP4	RTX 5090, NVFP4	1,540	1,408	1,490	1,465
DiffusionGemma-26B-A4B-it-ORL-NVFP4	B200, NVFP4	4,214	3,805	3,949	3,898
DiffusionGemma-26B-A4B-it-ORL-NVFP4	8× B200, NVFP4	7,708	7,123	7,026	6,871

The 8× B200 and RTX 5090 rows use generation seed 1234.

Usage

Plain transformers generation does not run the paper's decoding. Two engines do: dlmrt, our inference engine, and the ORL repository's patched vLLM (evaluation/inference/vllm/patches) with its recorded sampler, which the paper's evaluation uses.

dlmrt

dlmrt, our inference engine, decodes one request at a time with the paper's decoder (role="orl") and measures the paper's throughput; see its README for installation, tensor parallelism and the commands of every paper cell.

from dlmrt import LLM

llm = LLM("yifanyu/DiffusionGemma-26B-A4B-it-ORL", role="orl")
out = llm.generate("Write a Python function that reverses a linked list.", max_new_tokens=2048)
print(out.text)
print(f"{out.request_tps:.0f} tokens/s")

python -m dlmrt serve --model yifanyu/DiffusionGemma-26B-A4B-it-ORL --role orl --port 18080
curl -N localhost:18080/generate -d '{"messages": [{"role": "user", "content": "Hi!"}], "max_new_tokens": 512}'

The paper's evaluation pipeline

Evaluate it with the paper's protocol (GSM8K, MATH-500, HumanEval, MBPP) using the repository's evaluation pipeline; DiffusionGemma generation is not bit-deterministic, so results vary slightly between runs:

git clone https://github.com/Introspective-Diffusion/ORL orl && cd orl/evaluation
export ORL_EVAL_WORK=/path/to/eval-work
bash scripts/build_images.sh && bash scripts/prepare.sh
hf download yifanyu/DiffusionGemma-26B-A4B-it-ORL --local-dir $ORL_EVAL_WORK/hf/DiffusionGemma-26B-A4B-it-ORL
bash scripts/main_table.sh dg-orl $ORL_EVAL_WORK/hf/DiffusionGemma-26B-A4B-it-ORL

Serve it as the evaluation does (one 80 GB GPU; an OpenAI-compatible server on 127.0.0.1:8000, model name dg, with the GSM8K context length and sampler):

apptainer exec --nv -B "$PWD,$ORL_EVAL_WORK" --env PYTHONPATH=$PWD $ORL_EVAL_WORK/images/vllm-dg-orl.sif \
  python3 -m runtime.vllm.serve --model-root $ORL_EVAL_WORK/hf/DiffusionGemma-26B-A4B-it-ORL --role orl \
    --benchmark gsm8k --out $ORL_EVAL_WORK/dg-server --execute

Training

ORL trains on fresh, verified trajectories of the current model, replays the exact residual inputs each denoising step consumed, and adds adjacent entropy contraction (AEC) so that refinement gets faster, not only more accurate. This checkpoint is step 900 of the paper run on a mixture of 819 GSM8K, 819 MATH and 819 Python problems, on two nodes of 8 × H100 (actor and rollout). training/scripts/reproduce_diffusion_gemma.sh in the code repository reruns the recipe; MoE kernels accumulate in a nondeterministic order, so reruns differ slightly.

Checkpoint identity

Paper role dg-orl, training step 900. checkpoint_manifest.json records the hashes of the weights, tokenizer and configuration, and paper_decoding.json the engine and benchmark-specific decoding settings, which downloading the model does not apply.

Related Models

Model	Hugging Face	Description
DiffusionGemma-26B-A4B	google/diffusiongemma-26B-A4B-it	Start model of this checkpoint
QwenDLM-4B	yifanyu/QwenDLM-4B	Qwen3-4B-Thinking-2507 converted to block diffusion; start model of QwenDLM-4B-ORL
QwenDLM-4B-ORL	yifanyu/QwenDLM-4B-ORL	ORL on QwenDLM-4B
LLaDA2.1-mini-ORL	yifanyu/LLaDA2.1-mini-ORL	ORL on LLaDA-2.1-mini
LLaDA2.1-mini-ORL-NVFP4	yifanyu/LLaDA2.1-mini-ORL-NVFP4	NVFP4 quantization of LLaDA2.1-mini-ORL
DiffusionGemma-26B-A4B-it-ORL	yifanyu/DiffusionGemma-26B-A4B-it-ORL	This model
DiffusionGemma-26B-A4B-it-ORL-NVFP4	yifanyu/DiffusionGemma-26B-A4B-it-ORL-NVFP4	NVFP4 quantization of DiffusionGemma-26B-A4B-it-ORL
ORL-training-data	yifanyu/ORL-training-data	Dataset of the three ORL training runs

Citation

@misc{yu2026orl,
  title         = {On-Policy Residual Learning for Diffusion Language Models},
  author        = {Yu, Yifan and Jian, Yuqing and Li, Jisen and Hu, Yuezhou and Zhang, Tianyi and Wu, Xiaoxia and
                   Zhu, Rui-Jie and Wang, Junxiong and Zhou, Zhongzhu and Lai, Fan and Zou, James and Xu, Chenfeng},
  year          = {2026},
  eprint        = {2610.20687},
  archivePrefix = {arXiv},
  primaryClass  = {cs.LG},
  url           = {https://arxiv.org/abs/2610.20687}
}