FitMyGPU
Back to calculator

Qwen

Qwen 3 4B

Dense Qwen3 release with switchable thinking modes, stronger reasoning, and 131K extended-context support through YaRN.

Overview and architecture

What it is

Company

Qwen

Family

Qwen

Release date

Apr 27, 2025

Architecture

Dense decoder-only transformer

License

Apache 2.0

Modality

Text

Context window

131,072

Total params

4B

Active params

Dense model

Layers

36

Hidden size

2,560

Attention heads

32

KV heads

8

KV-bearing layers

36

Research highlight

What improved

Thinking-mode switch

The 4B model keeps the family’s dual-mode design, letting one checkpoint cover deeper reasoning and faster everyday dialogue.

Reasoning and agent uplift

Qwen emphasizes stronger reasoning, tool use, and agent-task performance over earlier Qwen2.5 instruct models.

Extended context with YaRN

The model is published with 32K native context and 131K support through YaRN, which is part of its practical release story.

Training and release context

How it was released

Family release

Qwen3 is released as a dense and MoE model family centered on switching between thinking and non-thinking modes within the same model.

Training stage

Qwen describes the release as a pretraining plus post-training model rather than a small instruction-only adaptation.

Context packaging

The 4B model is published with 32K native context, and the larger dense variants explicitly extend to 131K with YaRN.

Where it is strong

Where it is strong

Thinking and non-thinking use

The 4B release is built to switch between deeper reasoning mode and faster general dialogue mode without changing models.

Agent workflows

Qwen positions the family for tool use and agent-style tasks in both thinking and non-thinking modes.

Multilingual assistant work

The family is published with support for 100+ languages and dialects, making it a broad multilingual assistant line rather than a narrow specialist release.

Memory behavior

What dominates VRAM

This is still a dense model, so weights set the floor; the main extra lever is that longer contexts can extend to 131K when YaRN-style serving is used.

Sources

Where this page is grounded