AI Tools.

Search

text generation by facebook

opt-125m

OPT-125M is the smallest model in Meta's Open Pretrained Transformer series, a 125-million-parameter decoder-only LLM trained on a dataset comparable to GPT-3's training mix. Released as part of Meta's effort to make large language model weights accessible for research. At 125M parameters it is primarily used for prototyping, educational purposes, and compute-constrained environments.

Summary text generated by an automated pipeline from the model card · Not individually reviewed or run by us · How this page is made

From the model card

Fields below are copied from the tags and counters on the HuggingFace repository facebook/opt-125m at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.

Publisher (HF namespace)
facebook
Pipeline tag
text-generation
Library
Transformers
Framework tags
PyTorch, TensorFlow, JAX
License tag
other — read the license file in the repo before relying on it
Language tags
English (en)
Papers cited
arXiv:2205.01068, arXiv:2005.14165
Downloads (HF counter at last fetch)
9,676,549
Likes (HF counter at last fetch)
295
Model card
https://huggingface.co/facebook/opt-125m

Use cases

  • Lightweight text generation for prototyping and educational contexts
  • Minimal-resource LLM deployment on CPU-only machines
  • Research baseline for small LM behavior analysis
  • Fine-tuning starting point for domain-specific small generative models
  • Embedding extraction via hidden states when embedding models are unavailable

Pros

  • Tiny footprint — 125M params runs on virtually any hardware
  • Multi-framework support (PyTorch, TF, JAX)
  • Text-generation-inference compatible
  • Useful baseline for LLM scaling research

Cons

  • OPT license ('other') is not Apache/MIT — restricts some commercial uses
  • Severely outperformed by modern small LLMs (Qwen3-0.6B, Phi-3.5-mini) released since OPT
  • 125M parameters produce low quality generation on complex tasks
  • No instruction tuning — raw completion model requires careful prompting
  • Knowledge is dated; model released 2022 with earlier training cutoff

Tags

transformerspytorchtfjaxopttext-generationenarxiv:2205.01068arxiv:2005.14165license:othertext-generation-inferenceregion:usdeploy:azure