Qwen3-4B-Instruct-2507: Enhanced Instruction-Following LLM

Qwen3-4B-Instruct-2507 is an updated 4.0 billion parameter causal language model from Qwen, designed for superior instruction following and long-context understanding. It features a remarkable native context length of 262,144 tokens, making it highly capable for tasks requiring extensive information processing.

Key Capabilities

General Performance: Significant improvements across instruction following, logical reasoning, text comprehension, mathematics, science, and coding.
Multilingual Knowledge: Substantial gains in long-tail knowledge coverage across various languages.
User Alignment: Markedly better alignment with user preferences for subjective and open-ended tasks, leading to more helpful and higher-quality text generation.
Tool Usage: Enhanced capabilities in tool usage, with recommendations to leverage Qwen-Agent for optimal agentic abilities.
Non-Thinking Mode: This model exclusively operates in a "non-thinking mode," simplifying usage by not generating <think></think> blocks.

Performance Highlights

Benchmarks show Qwen3-4B-Instruct-2507 achieving leading scores in several categories compared to its predecessor and other models, including:

Knowledge: MMLU-Pro (69.6), GPQA (62.0)
Reasoning: AIME25 (47.4), HMMT25 (31.0), ZebraLogic (80.2)
Coding: LiveCodeBench v6 (35.1), MultiPL-E (76.8)
Alignment: Creative Writing v3 (83.5), WritingBench (83.4)
Agent: BFCL-v3 (61.9), TAU1-Retail (48.7)

Recommended Use Cases

This model is ideal for applications requiring:

Advanced instruction following and complex reasoning.
Processing and generating content based on very long contexts.
Multilingual applications and tasks demanding broad knowledge.
Agentic workflows and tool-calling functionalities.