What are the hardware requirements for Opt 350M?

Opt 350M requires approximately 0.7 GB of VRAM and supports a context window of 2K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Opt 350M and how large is it?

Opt 350M is developed by facebook, a model with 350m parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Opt 350M perform on standard benchmarks?

Opt 350M has the following published scores: MMLU 26.02. Compare against reference models on this page or on the LLM Explorer leaderboards.

Opt 350M by facebook — VRAM 0.7GB, 2K context

Name: Opt 350M
Author: facebook

Opt 350M is an open-source language model by facebook. Features: 350m LLM, VRAM: 0.7GB, Context: 2K, License: other, LLM Explorer Score: 0.13, Arc: 23.6, HellaSwag: 36.7, MMLU: 26, GSM8K: 0.3.

Arxiv:2005.14165 Arxiv:2205.01068 Deploy:azure En Jax Opt Pytorch Region:us Tf

Model Card on HF 🤗: https://huggingface.co/facebook/opt-350m

Opt 350M Parameters and Internals

Model Type

text generation, decoder-only model

Use Cases

Areas:

Research, Text Generation

Primary Use Cases:

prompting for evaluation, text generation

Limitations:

bias, toxicity, generation diversity issues, hallucination

Considerations:

Fine-tuned models will inherit biases from the base model.

Supported Languages

English (Predominantly)

Training Details

Data Sources:

BookCorpus, CC-Stories, The Pile including subsets like Pile-CC, OpenWebText2, USPTO, Project Gutenberg, OpenSubtitles, Wikipedia, DM Mathematics, HackerNews, Pushshift.io Reddit, CCNewsV2

Data Volume:

180B tokens

Methodology:

Pretrained using a causal language modeling (CLM) objective

Context Length:

2048

Training Time:

~33 days

Hardware Used:

992 *80GB A100 GPUs

Model Architecture:

Causal language modeling

Safety Evaluation

Risk Categories:

bias, toxicity, safety issues

Ethical Considerations:

The training data contains unfiltered content, which is not neutral leading to biased outputs.

Responsible Ai Considerations

Transparency:

Data sources and limitations mentioned, but specifics of transparency not detailed.

Input Output

Accepted Modalities:

text

Output Format:

Generated text

LLM Name	Opt 350M
Repository 🤗	https://huggingface.co/facebook/opt-350m
Model Size	350m
Required VRAM	0.7 GB
Updated	2026-07-17
Maintainer	facebook
Model Type	opt
Model Files	0.7 GB
Supported Languages	en
Model Architecture	OPTForCausalLM
License	other
Context Length	2048
Model Max Length	2048
Transformers Version	4.20.0.dev0
Beginning of Sentence Token	</s>
End of Sentence Token	</s>
Unk Token	</s>
Vocabulary Size	50272
Torch Data Type	float16
Activation Function	relu
Errors	replace

Rank the Opt 350M capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Opt 350M

Best Alternatives	Context / RAM	Downloads	Likes
Opt Mini Dataset 0	2K / 0.7 GB	8	0
Facebook Opt 350M SFT Korz14	2K / 0.7 GB	10	0
Temp Model Sft	2K / 1.3 GB	7	0
Gpt350 Chat S V0	2K / 0.7 GB	6	0
Gpt350 Chat S V0 1	2K / 0.7 GB	5	0
Dadjokes Tuned Opt	2K / 1.3 GB	8	2
Pygmalion 350M	2K / 1.3 GB	348	54
Rockyalquimista888	2K / 1.3 GB	8	0
Remedycure	2K / 1.3 GB	6	0
Sunday	2K / 1.3 GB	6	0

Note: green Score (e.g. "73.2") means that the model is better than facebook/opt-350m.