Instructions to use goodgoals/Neuron-NxT-Step with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use goodgoals/Neuron-NxT-Step with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="goodgoals/Neuron-NxT-Step")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("goodgoals/Neuron-NxT-Step", dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use goodgoals/Neuron-NxT-Step with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "goodgoals/Neuron-NxT-Step" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "goodgoals/Neuron-NxT-Step", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/goodgoals/Neuron-NxT-Step
- SGLang
How to use goodgoals/Neuron-NxT-Step 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 "goodgoals/Neuron-NxT-Step" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "goodgoals/Neuron-NxT-Step", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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 "goodgoals/Neuron-NxT-Step" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "goodgoals/Neuron-NxT-Step", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use goodgoals/Neuron-NxT-Step with Docker Model Runner:
docker model run hf.co/goodgoals/Neuron-NxT-Step
Welcome to NeuronNXTStep: Introduction
NeuronNXTStep is our newest, smartest reasoning Language Model yet. It has much better benchmarks compared to our older models, for instance, the Natarajan Response Engine v1.02, which was our most popular model yet, was based on GPT-OSS-20b, which was an exellent model, but unmatched compared to NeuronNxTStep whose base is the exellent Qwen QwQ 32b. NeuronNxtStep has exellent benchmarks, rivaling heavier, more resource hungry LLMs like OpenAI's o1 mini, Deepseek R1 671b, Llama models, and Deepseek R1 Distills! It brings an end to the dillema between smarter or efficient! For a more specific overview, here are the benchmarks:
Speciality
NeuronNxTStep also addresses a big issue in the constantly growing world of LLMs. As LLMs get smarter and are forced to get the right answer, they find ways to manipulate and cheat. That also includes the manipulation of humans which introduces new dangers. To reduce risk of safety misalignment, we trained this model using the effective UCSC-VLAA/STAR1 dataset. This measure helps proof the LLM from misalignment due to them constantly growing and finding more ways to manipulate humans.
Notes
- The model files given in this repository are the model tensors for the NeuronNxTStep LLM.
- This model uses the Apache Licence 2.0, and therefore grants you permission to fine tune this model, run it locally, comercialize, etc.
- This model may have a slight accuracy loss as it uses 4-bit quantanization
- The LLM was fine tuned with Unsloth, the memory efficient tool that allows you to fine tune bigger LMs on less powerful GPUs.
- Gradient Checkpointing was already enabled during the making of this model. To ensure a smooth experience, don't re-enable gradient checkpointing, otherwise you may hit