Instructions to use OpenOneRec/OneReason-8B-pretrain-competition with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OpenOneRec/OneReason-8B-pretrain-competition with Transformers:

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

pipe = pipeline("text-generation", model="OpenOneRec/OneReason-8B-pretrain-competition")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("OpenOneRec/OneReason-8B-pretrain-competition")
model = AutoModelForCausalLM.from_pretrained("OpenOneRec/OneReason-8B-pretrain-competition", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use OpenOneRec/OneReason-8B-pretrain-competition with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OpenOneRec/OneReason-8B-pretrain-competition"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OpenOneRec/OneReason-8B-pretrain-competition",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/OpenOneRec/OneReason-8B-pretrain-competition

SGLang

How to use OpenOneRec/OneReason-8B-pretrain-competition 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 "OpenOneRec/OneReason-8B-pretrain-competition" \
    --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": "OpenOneRec/OneReason-8B-pretrain-competition",
		"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 "OpenOneRec/OneReason-8B-pretrain-competition" \
        --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": "OpenOneRec/OneReason-8B-pretrain-competition",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use OpenOneRec/OneReason-8B-pretrain-competition with Docker Model Runner:
```
docker model run hf.co/OpenOneRec/OneReason-8B-pretrain-competition
```

OpenOneRec commited on Jul 29

Commit

8e08cbf

verified ·

1 Parent(s): e11197a

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +17 -12

README.md CHANGED Viewed

@@ -11,9 +11,10 @@ tags:
 - itemic-token
 - qwen3
 - pretraining
 ---
-# OneReason
 Reasoning Foundation Models for Generative Recommendation
@@ -35,23 +36,25 @@ The OneReason training stack contains three stages:
 - **Supervised Fine-Tuning (SFT):** teaches recommendation cognition with coarse-to-fine Chain-of-Thought (CoT) traces over user profiles, behavior histories, and itemic-token evidence.
 - **Reinforcement Learning (RL):** uses a specialize-then-unify recipe to improve thinking-mode recommendation while balancing performance across multiple recommendation domains.
-This repository currently releases the **OneReason-0.8B Pretrain checkpoint**. We will continue to release OneReason-0.8B SFT/RL checkpoints and the OneReason-8B series.
 ## News
 - **[2026.06]** OneReason-0.8B Pretrain checkpoint is released.
-- **Coming soon:** OneReason-0.8B SFT checkpoint.
-- **Coming soon:** OneReason-0.8B RL checkpoint.
-- **Coming soon:** OneReason-8B checkpoints.
 ## Model Zoo
 | Model | Stage | Parameters | Status | Description |
 |---|---:|---:|---|---|
 | OneReason-0.8B-Pretrain | Pre-training | 0.8B | Released | Foundation checkpoint after itemic-text alignment pre-training. Suitable for research, continued pre-training, and downstream SFT. |
 | OneReason-0.8B-SFT | SFT | 0.8B | Coming soon | Instruction-tuned checkpoint with recommendation perception, derivation, evolution, and recommendation supervision. |
 | OneReason-0.8B-RL | RL | 0.8B | Coming soon | Post-trained checkpoint optimized for recommendation-oriented reasoning. |
-| OneReason-8B | Pretrain/SFT/RL | 8B | Coming soon | Larger OneReason model family with stronger reasoning and recommendation performance. |
 ## Method Overview
@@ -109,9 +112,9 @@ OneReason is evaluated with **OneReason-Bench**, a reasoning-oriented recommenda
 ## Performance
-The released **OneReason-0.8B-Pretrain** checkpoint is the foundation checkpoint before SFT/RL. It is designed to provide strong itemic-token perception and a good initialization for downstream recommendation tuning.
-The tables below report the full OneReason-8B system results from the technical report. We will update this model card with checkpoint-specific numbers as the OneReason-0.8B SFT/RL and OneReason-8B checkpoints become available.
 <p align="center">
   <img src="assert/rader.png" alt="OneReason performance overview and thinking-mode gains" width="95%">
@@ -177,7 +180,7 @@ Load the model:
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
-model_name = "OpenOneRec/OneReason-0.8B-Pretrain"  # or the local path to this repository
 tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
@@ -299,12 +302,13 @@ Expected response (example output):
 ## Intended Use
-The OneReason-0.8B Pretrain checkpoint is intended for:
-- Research on generative recommendation and recommendation foundation models.
 - Continued pre-training or SFT on new recommendation domains.
 - Itemic-token perception studies, including item understanding and itemic-token grounding.
 - Building downstream recommendation models that combine user profiles, behavior histories, and itemic-token representations.
 For best recommendation reasoning performance, we recommend using future SFT/RL checkpoints once released, or fine-tuning this pretrain checkpoint on task-specific supervised data.
@@ -314,7 +318,8 @@ For best recommendation reasoning performance, we recommend using future SFT/RL
 - Direct recommendation quality depends on the itemic tokenizer, item catalog, user history format, and decoding strategy.
 - Generated itemic tokens must be validated against the target item catalog before being used as item IDs.
 - The model may generate invalid, stale, or unsupported itemic-token sequences if the prompt distribution differs significantly from training data.
-- The checkpoint is released for research and should not be used for high-stakes personalization without careful evaluation, filtering, and privacy review.
 ## Citation

 - itemic-token
 - qwen3
 - pretraining
+- competition
 ---
+# OneReason-8B-Pretrain (Competition)
 Reasoning Foundation Models for Generative Recommendation
 - **Supervised Fine-Tuning (SFT):** teaches recommendation cognition with coarse-to-fine Chain-of-Thought (CoT) traces over user profiles, behavior histories, and itemic-token evidence.
 - **Reinforcement Learning (RL):** uses a specialize-then-unify recipe to improve thinking-mode recommendation while balancing performance across multiple recommendation domains.
+This repository releases the **OneReason-8B-Pretrain (Competition)** checkpoint — the 8B foundation checkpoint after itemic-text alignment pre-training, prepared for the OneReason competition track. The 0.8B pretrain checkpoint is released in a sibling repository.
 ## News
+- **[2026.07]** OneReason-8B-Pretrain (Competition) checkpoint is released.
 - **[2026.06]** OneReason-0.8B Pretrain checkpoint is released.
+- **Coming soon:** OneReason-0.8B / 8B SFT checkpoints.
+- **Coming soon:** OneReason-0.8B / 8B RL checkpoints.
 ## Model Zoo
 | Model | Stage | Parameters | Status | Description |
 |---|---:|---:|---|---|
 | OneReason-0.8B-Pretrain | Pre-training | 0.8B | Released | Foundation checkpoint after itemic-text alignment pre-training. Suitable for research, continued pre-training, and downstream SFT. |
+| OneReason-8B-Pretrain (Competition) | Pre-training | 8B | Released | Larger foundation checkpoint with stronger itemic-token perception and broader recommendation coverage. This is the competition release. |
 | OneReason-0.8B-SFT | SFT | 0.8B | Coming soon | Instruction-tuned checkpoint with recommendation perception, derivation, evolution, and recommendation supervision. |
 | OneReason-0.8B-RL | RL | 0.8B | Coming soon | Post-trained checkpoint optimized for recommendation-oriented reasoning. |
+| OneReason-8B-SFT | SFT | 8B | Coming soon | Instruction-tuned 8B checkpoint. |
+| OneReason-8B-RL | RL | 8B | Coming soon | Post-trained 8B checkpoint optimized for recommendation-oriented reasoning. |
 ## Method Overview
 ## Performance
+The released **OneReason-8B-Pretrain (Competition)** checkpoint is the 8B foundation checkpoint before SFT/RL. It provides stronger itemic-token perception and broader recommendation coverage than the 0.8B release, and serves as a strong initialization for downstream recommendation tuning.
+The tables below report the full OneReason-8B system results from the technical report. The pretrain checkpoint itself is evaluated on R0 perception and grounding tasks; the SFT/RL rows are included to show the full system performance that this pretrain checkpoint enables.
 <p align="center">
   <img src="assert/rader.png" alt="OneReason performance overview and thinking-mode gains" width="95%">
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
+model_name = "OpenOneRec/OneReason-8B-pretrain-competition"  # or the local path to this repository
 tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
 ## Intended Use
+The OneReason-8B-Pretrain (Competition) checkpoint is intended for:
+- Research on generative recommendation and recommendation foundation models at the 8B scale.
 - Continued pre-training or SFT on new recommendation domains.
 - Itemic-token perception studies, including item understanding and itemic-token grounding.
 - Building downstream recommendation models that combine user profiles, behavior histories, and itemic-token representations.
+- Competition baselines that require a strong, open recommendation foundation checkpoint.
 For best recommendation reasoning performance, we recommend using future SFT/RL checkpoints once released, or fine-tuning this pretrain checkpoint on task-specific supervised data.
 - Direct recommendation quality depends on the itemic tokenizer, item catalog, user history format, and decoding strategy.
 - Generated itemic tokens must be validated against the target item catalog before being used as item IDs.
 - The model may generate invalid, stale, or unsupported itemic-token sequences if the prompt distribution differs significantly from training data.
+- The 8B checkpoint requires more memory and compute than the 0.8B release; make sure your environment has sufficient GPU resources before loading.
+- The checkpoint is released for research and competition use, and should not be used for high-stakes personalization without careful evaluation, filtering, and privacy review.
 ## Citation