Stable Beluga 2 uses a Llama 2 70B model fine-tuned on an Orca-style dataset and offers a research preview chat.
Dataset
Stable Beluga
Stable Beluga 1 is a Llama 65B model fine-tuned on an Orca-style dataset; its page provides delta weights.
About Stable Beluga
Stable Beluga 1 is a 65-billion parameter text generation model fine-tuned on an Orca-style dataset, hosted on the Hugging Face platform. It is built on the LLaMA foundation model and released under a CC BY-NC 4.0 license for non-commercial use.
The model is distributed as delta weights only, so it cannot be used standalone. Users must combine these weights with original LLaMA 65B model weights to create a functional version of Stable Beluga 1. A provided apply_delta.py script automates this merging process, and sample code is included for loading the final model and running text generation tasks. It is designed for instruction-following use cases, with example prompts including creative writing, factual questions, and concept explanations.
Key features
- Fine-tuned on Orca-style dataset for instruction following
- Provided as delta weights with automated merge script
- Compatible with Hugging Face Transformers library
- Supports English language text generation tasks
- Released under CC BY-NC 4.0 non-commercial license
Best for: It is best suited for NLP researchers and developers working on non-commercial instruction-following text generation projects.
Good for
- Built on LLaMA 65B foundation for strong base performance
- Includes pre-built script to simplify delta weight merging
- Integrates with standard Hugging Face Transformers workflows
Watch out for
- Requires separate access to original LLaMA 65B weights to use
- Restricted to non-commercial use under CC BY-NC 4.0 license
- No hosted inference providers available on its Hugging Face page
Stable Beluga at a glance
- Official website
- huggingface.co/stabilityai/StableBeluga1-Delta
- Category
- AI & Machine Learning › Models & ML Research
- Type
- Dataset
- Platforms
- Web
- Last checked
- Oct 5, 2026 · Online when checked
- Listed
- Oct 5, 2026
Similar data sources
Other listings in Models & ML Research with the most in common.
Popular in Models & ML Research
Well-known listings from the same topic.
Falcon LLMfalconllm.tii.aeERNIEernie.baidu.comTokenVersetoken-verse.github.iowTAObittensor.comAlpacacrfm.stanford.edu/2023/03/13/alpaca.htmlSmolVLAhuggingface.co/blog/smolvla
Stable Beluga FAQ
What is Stable Beluga 1?
Stable Beluga 1 is a 65B parameter text generation model fine-tuned on an Orca-style dataset, built on the LLaMA foundation model.
How do I use Stable Beluga 1?
You must merge the provided delta weights with original LLaMA 65B weights using the included apply_delta.py script, then load the combined model with Hugging Face Transformers.
What license covers Stable Beluga 1?
It is released under a CC BY-NC 4.0 license, which permits only non-commercial use.
Is hosted inference available for Stable Beluga 1?
No, no inference providers are listed for the model on its Hugging Face page.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.