---
title: Qwen/Qwen3-4B-Instruct-2507
description: Qwen/Qwen3-4B-Instruct-2507 — Qwen3 model with 4.0B parameters.
canonical_url: https://superlinked.com/models/qwen-qwen3-4b-instruct-2507
last_updated: 2026-06-29
---

# Qwen/Qwen3-4B-Instruct-2507

Source: [Qwen/Qwen3-4B-Instruct-2507 on HuggingFace](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507)

## Overview

| Field | Value |
|-------|-------|
| Architecture | Qwen3 |
| Parameters | 4.0B |
| Tasks | Generate |
| Outputs | Text |
| Context window | 32,768 tokens |
| Max output | 4,096 tokens |
| Capabilities | tool calling, constrained output (JSON Schema, Regex), streaming, code, SQL |
| License | apache-2.0 |
| Inputs | text |

## Benchmarks

### HumanEval

Domain: code · Task: generation · Language: en

**Quality:** code pass at 1: 0.8659

### MBPP

Domain: code · Task: generation · Language: en

**Quality:** code pass at 1: 0.7400

### Spider

Domain: sql · Task: generation · Language: en

**Quality:** sql execution accuracy: 0.6900

### BFCL (simple)

Domain: tools · Task: generation · Language: en

**Quality:** tool call ast: 0.9375

### BFCL (multiple)

Domain: tools · Task: generation · Language: en

**Quality:** tool call ast: 0.9200

### CaseHOLD

Domain: legal · Task: generation · Language: en

**Quality:** accuracy: 0.6033

**Performance (RTX-PRO-6000 b1 c4):** Corpus 441 tok/s · Corpus p50 607.3ms

### GPQA Diamond

Domain: scientific · Task: generation · Language: en

**Quality:** accuracy: 0.4444

**Performance (RTX-PRO-6000 b1 c4):** Corpus 495 tok/s · Corpus p50 1.2s

### MedQA

Domain: medical · Task: generation · Language: en

**Quality:** accuracy: 0.5700

**Performance (RTX-PRO-6000 b1 c4):** Corpus 475 tok/s · Corpus p50 545.4ms

### MMLU-Pro

Domain: general · Task: generation · Language: en

**Quality:** accuracy: 0.5333

**Performance (RTX-PRO-6000 b1 c4):** Corpus 468 tok/s · Corpus p50 446.4ms