Qwen/Qwen3.5-4B is a multimodal LLM with native vision capabilities. It excels in general-purpose dialogue, instruction following, coding, mathematics, multilingual tasks, and vision-language understanding, with strong performance on reasoning, agents, and visual understanding benchmarks. The 4B parameter scale balances capability and efficiency for a wide range of deployment scenarios.

Key features include thinking mode by default for complex reasoning, a unified vision-language foundation with early fusion training, support for 201 languages and dialects, long-context processing up to 262,144 tokens (extensible to 1,010,000 with YaRN), image and video understanding, tool calling for agentic use cases, and a hybrid architecture using Gated Delta Networks with sparse Mixture-of-Experts.

Metric	Value
Parameter Count	4 billion
Mixture of Experts	No
Context Length	262,144 tokens
Multilingual	Yes
Quantized*	No

*Quantization is specific to the inference provider and the model may be offered with different quantization levels by other providers.