Instructions to use ai21labs/AI21-Jamba-Reasoning-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ai21labs/AI21-Jamba-Reasoning-3B with Transformers:

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

pipe = pipeline("text-generation", model="ai21labs/AI21-Jamba-Reasoning-3B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ai21labs/AI21-Jamba-Reasoning-3B")
model = AutoModelForCausalLM.from_pretrained("ai21labs/AI21-Jamba-Reasoning-3B", 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 ai21labs/AI21-Jamba-Reasoning-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ai21labs/AI21-Jamba-Reasoning-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ai21labs/AI21-Jamba-Reasoning-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ai21labs/AI21-Jamba-Reasoning-3B

SGLang

How to use ai21labs/AI21-Jamba-Reasoning-3B 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 "ai21labs/AI21-Jamba-Reasoning-3B" \
    --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": "ai21labs/AI21-Jamba-Reasoning-3B",
		"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 "ai21labs/AI21-Jamba-Reasoning-3B" \
        --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": "ai21labs/AI21-Jamba-Reasoning-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ai21labs/AI21-Jamba-Reasoning-3B with Docker Model Runner:
```
docker model run hf.co/ai21labs/AI21-Jamba-Reasoning-3B
```

MLX optimization

by do-me - opened Oct 13, 2025

Discussion

do-me

Oct 13, 2025

•

edited Oct 13, 2025

Hey thanks a lot for this dev effort - awesome to see models with new architectures performing this well!
I saw that you used a Mac in the benchmarks. Are you by chance interested in optimizing the model for MLX?
There is https://huggingface.co/mlx-community but jamba architecture is not yet supported. Would you be interested in contributing to https://github.com/ml-explore/mlx-lm?
I guess there could be significant speed-ups.

Personally I am currently processing large text dumps for a research prioject and looking for a model that is:

small & fast
with high quality outputs
supporting long-contexts
high batch throughput
optimized for MLX

So far you model ticks most bullets for me me. If you optimized it for MLX, I could use batch generation easily (https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/examples/batch_generate_response.py).

Best
Dominik

Update: opened this issue in mlx-lm https://github.com/ml-explore/mlx-lm/issues/551

Goekdeniz-Guelmez

Oct 14, 2025

•

edited Oct 14, 2025

Hey Dominik,

I created a PR last week, I just finished it so it will be merged soon.

Cheers
Gökdeniz

do-me

Oct 14, 2025

Awesome, thanks a lot!

Goekdeniz-Guelmez

Oct 15, 2025

the PR has been merged, the next MLX-LM release will support Jamba.

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