Qwen3-1.7B-Base

Qwen3-1.7B-Base

Continues text and code as a 1.7B base language model pre-trained on 119 languages. For developers who want a small model to fine-tune or test.

Made by Qwen on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Qwen3-1.7B-Base is a small base language model from the Qwen3 family. It has 1.7 billion parameters and sits at the pretraining stage, so it continues text rather than following chat instructions. Its pretraining covered 36 trillion tokens in 119 languages.

Who it suits: It suits developers and researchers who want a compact base model to fine-tune, quantize or test ideas on. Hugging Face lists 56 quantized and 504 fine-tuned versions made by other people.

Industry | Who uses it | How they use it | Example
AI research | ML researchers | Test training ideas on a small base model | Comparing two fine-tuning recipes on the same starting point
Software development | Application developers | Fine-tune it for a narrow text task | Adapting it to complete text in one product
Model distribution | Packagers | Make smaller quantized copies | Shrinking the weights from BF16 to 4-bit

You can skip it when you need a model that follows instructions out of the box, since this is the base version.

Where to use it:

• Python with the transformers library
• Windows, macOS or Linux, once the files are downloaded
• Your own fine-tuning and quantizing workflows

Use case | Who | What they get back
Text continuation | Developers | Text that carries on from the words you give it
Fine-tuning base | ML engineers | A starting point to train for one task
Quantized copies | Packagers | A smaller version for lighter hardware

How to set it up:

• Python on Windows, macOS or Linux
• The huggingface_hub package, to download the files
• The latest transformers version, since older ones fail on this model
• Room for about 3.4 GB of BF16 weights

1. Install the Hugging Face hub package.
2. Download the model files to your computer.
3. Keep transformers up to date.

pip install -U huggingface_hub
hf download Qwen/Qwen3-1.7B-Base

With a transformers version older than 4.51.0, loading stops with this error.

KeyError: 'qwen3'

How to use it: Day to day you load the model with transformers and give it some text to continue. The card shows no chat prompt or code sample, and points to its blog and documentation for more.

Tool | What it does
hf download | Fetches the model files to your computer
transformers | Loads the model, in its latest version

Example: You give the model the first sentences of a paragraph. It carries on writing from there as plain text completion. It does not answer you as a chat assistant would, because it is a base model.

For newcomers and experts: If you are new, download the files with the two commands above and read the card. Then load the model with transformers and give it a half-written sentence to finish.

If you know the field, this is a causal language model with 28 layers, 1.4 billion non-embedding parameters and grouped query attention with 16 query heads and 8 key-value heads. Training ran in three stages, from broad language modeling to reasoning and then longer sequences. Context length is 32,768.

You give it the start of some text
The model reads it through 28 layers
It predicts the next tokens
The text continues

Good to know:

Title: Attention heads
Query heads: 16 heads
Key-value heads: 8 heads

• Base model: the card describes the pretraining stage only
• Context: 32,768 tokens
• Size: about 3.4 GB in BF16, or about 860 MB at 4-bit as an estimate
• License: apache-2.0
• Cost to run: none named, it runs on your own hardware
• Data it reaches: the text you give it, and the model files you download from Hugging Face
• Not for: following chat instructions, since the card covers only the pretrained base

On Hugging Face, 13 November 2026

• Task: text generation
• Library: transformers
• Parameters: 1.7B
• Weights: about 3.4 GB in BF16; about 860 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 599,054
• Likes: 101
• License: apache-2.0
• Quantized versions: 56
• Fine-tunes: 504
• Tags: transformers, safetensors, qwen3, text-generation, conversational

Links

• Model: https://huggingface.co/Qwen/Qwen3-1.7B-Base
• Files: https://huggingface.co/Qwen/Qwen3-1.7B-Base/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Qwen/Qwen3-1.7B-Base

Then follow the usage in its model card. Nothing in this kit runs by itself.