Instructions to use allenai/Olmo-3.1-32B-Think with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use allenai/Olmo-3.1-32B-Think with Transformers:

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

pipe = pipeline("text-generation", model="allenai/Olmo-3.1-32B-Think")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("allenai/Olmo-3.1-32B-Think")
model = AutoModelForCausalLM.from_pretrained("allenai/Olmo-3.1-32B-Think", 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 allenai/Olmo-3.1-32B-Think with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "allenai/Olmo-3.1-32B-Think"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "allenai/Olmo-3.1-32B-Think",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/allenai/Olmo-3.1-32B-Think

SGLang

How to use allenai/Olmo-3.1-32B-Think 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 "allenai/Olmo-3.1-32B-Think" \
    --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": "allenai/Olmo-3.1-32B-Think",
		"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 "allenai/Olmo-3.1-32B-Think" \
        --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": "allenai/Olmo-3.1-32B-Think",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use allenai/Olmo-3.1-32B-Think with Docker Model Runner:
```
docker model run hf.co/allenai/Olmo-3.1-32B-Think
```

Chat template - no <think> when applying chat template to messages including assistant responses, is this intentional?

by kdu4108 - opened Feb 4

Discussion

kdu4108

Feb 4

Hi,

Let's say I generated text using Olmo 3 Think. Then it might give me something like "Okay, let's think, ...., \boxed{1}.</think>\n\nThe final answer is \boxed{1}.". Notably, the opening <think>tag will NOT be there because the chat template for tokenizer.apply_chat_template([{"role": "user", "content": "What's 2-1?"}], add_generation_prompt=True), will already include the <think> token in the input prompt to the model.

However, let’s say I wanted to then get the embeddings of the full prompt including the model’s response, or continue the chat, or whatever: when running
tokenizer.apply_chat_template([{"role": "user", "content": "What's 2-1?"}, {"role": "assistant", "content":"Okay, let's think, ...., \boxed{1}.</think>\n\nThe final answer is \boxed{1}."}], add_generation_prompt=False), you would get:
'<|im_start|>system\nYou are OLMo, a helpful function-calling AI assistant built by Ai2. Your date cutoff is November 2024, and your model weights are available at https://huggingface.co/allenai. You do not currently have access to any functions. <functions></functions><|im_end|>\n<|im_start|>user\nhello there<|im_end|>\n<|im_start|>assistant\nOkay, let's think, ...., \boxed{1}.</think>\n\nThe final answer is \boxed{1}.<|endoftext|>’.

(and if you set it to True, then you get same but with the <|im_start|>assistant\n<think> at the end which is also not what I want.)

Notably, the <think> is now missing, because the chat template does not include it for the assistant except if its starting the assistant generation! Is this intentional?

Thanks!

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment