What are the hardware requirements for GPT Sw3 1.3B Instruct?

GPT Sw3 1.3B Instruct requires approximately 5.5 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed GPT Sw3 1.3B Instruct and how large is it?

GPT Sw3 1.3B Instruct is developed by AI-Sweden-Models, a model with 1.3b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does GPT Sw3 1.3B Instruct perform on standard benchmarks?

GPT Sw3 1.3B Instruct has the following published scores: MMLU 26.17. Compare against reference models on this page or on the LLM Explorer leaderboards.

GPT Sw3 1.3B Instruct by AI-Sweden-Models — VRAM 5.5GB

Name: GPT Sw3 1.3B Instruct
Author: AI-Sweden-Models

GPT Sw3 1.3B Instruct is an open-source language model by AI-Sweden-Models. Features: 1.3b LLM, VRAM: 5.5GB, License: other, Instruction-Based, LLM Explorer Score: 0.12, Arc: 31, HellaSwag: 51.4, MMLU: 26.2, GSM8K: 1.6.

Base model:ai-sweden-models/gp... Base model:finetune:ai-sweden-... Conversational Da Dataset:databricks/databricks-... Dataset:laion/oig Dataset:openassistant/oasst1 En Endpoints compatible Gpt2 Instruct Is No Pytorch Region:us Safetensors Sv

Model Card on HF 🤗: https://huggingface.co/AI-Sweden-Models/gpt-sw3-1.3b-instruct

GPT Sw3 1.3B Instruct Parameters and Internals

Model Type

language model, text generation

Use Cases

Areas:

Nordic NLP ecosystem

Applications:

Pre-release for research and evaluation of the capabilities of Large Language Models for the Nordic languages.

Primary Use Cases:

GPT-SW3 can generate coherent text in multiple languages and perform text tasks by casting them as text generation tasks.

Limitations:

Bias, Safety, Generation diversity issues, Hallucination, Overrepresentation/Underrepresentation of certain viewpoints, Stereotypes, May generate inappropriate content

Supported Languages

da (Unknown), sv (Unknown), no (Unknown), en (Unknown), is (Unknown)

Training Details

Data Sources:

laion/OIG, databricks/databricks-dolly-15k, OpenAssistant/oasst1

Data Volume:

320B tokens

Methodology:

Trained with the NeMo Megatron GPT implementation.

Model Architecture:

Decoder-only transformer language model.

Input Output

Input Format:

Raw text or instruction data in chat format.

Accepted Modalities:

text

Output Format:

Generated text

LLM Name	GPT Sw3 1.3B Instruct
Repository 🤗	https://huggingface.co/AI-Sweden-Models/gpt-sw3-1.3b-instruct
Base Model(s)	AI-Sweden-Models/gpt-sw3-1.3b AI-Sweden-Models/gpt-sw3-1.3b
Model Size	1.3b
Required VRAM	5.5 GB
Updated	2026-06-06
Maintainer	AI-Sweden-Models
Model Type	gpt2
Instruction-Based	Yes
Model Files	5.5 GB 5.5 GB
Supported Languages	da sv no en is
Model Architecture	GPT2LMHeadModel
License	other
Model Max Length	2048
Transformers Version	4.22.1
Tokenizer Class	GPTSw3Tokenizer
Vocabulary Size	64000
Torch Data Type	float32
Activation Function	gelu

Rank the GPT Sw3 1.3B Instruct 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