ByteDance
/

Ouro-2.6B-Thinking

Text Generation

looped-language-model

recurrent-depth

chain-of-thought

Model card Files Files and versions

Instructions to use ByteDance/Ouro-2.6B-Thinking with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ByteDance/Ouro-2.6B-Thinking with Transformers:

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

pipe = pipeline("text-generation", model="ByteDance/Ouro-2.6B-Thinking", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("ByteDance/Ouro-2.6B-Thinking", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

How to use ByteDance/Ouro-2.6B-Thinking with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ByteDance/Ouro-2.6B-Thinking"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ByteDance/Ouro-2.6B-Thinking",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ByteDance/Ouro-2.6B-Thinking

How to use ByteDance/Ouro-2.6B-Thinking 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 "ByteDance/Ouro-2.6B-Thinking" \
    --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": "ByteDance/Ouro-2.6B-Thinking",
		"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 "ByteDance/Ouro-2.6B-Thinking" \
        --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": "ByteDance/Ouro-2.6B-Thinking",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ByteDance/Ouro-2.6B-Thinking with Docker Model Runner:
```
docker model run hf.co/ByteDance/Ouro-2.6B-Thinking
```

Ouro-2.6B-Thinking

Commit History

Updated ids for bos_id, eos_id (#4)

f1edd81

sirorezka commited on Feb 26

Added 'pad_token_id'. (#5)

bed729a

sirorezka commited on Feb 26

rope_type='default' excluded from ROPE_INIT_FUNCTIONS in transfomers >=5.0 (#6)

d917c94

sirorezka commited on Feb 26

Fix bos/eos token IDs + add enable_thinking to chat template (#7)

2d13cf8

KristianS7 commited on Feb 26

Fix UniversalTransformerCache.get_mask_sizes for batched generation (#8)

3201be3

KristianS7 commited on Feb 26

fix BOS/EOS

85a074d
verified

ridger commited on Jan 18

Update README.md

dafed72
verified

ridger commited on Nov 16, 2025

Update modeling_ouro.py

1855d8a
verified

ridger commited on Nov 16, 2025

Update Ouro-2.6B-Thinking model card with paper and code links (#1)

a2d3a54
verified

nielsr HF Staff commited on Oct 31, 2025

Upload folder using huggingface_hub

d5d2e61
verified

ridger commited on Oct 29, 2025

Update README.md

007a3d0
verified

ridger commited on Oct 28, 2025

Upload folder using huggingface_hub

cfe6f63
verified

ridger commited on Oct 28, 2025

initial commit

18f36b9
verified

ridger commited on Oct 28, 2025