Qwen3-4B

Qwen3-4B represents the next level in the series. Built on a deeper architecture with 4 billion parameters, it features 36 layers and double the number of attention heads (32/8), delivering significant improvements in understanding and generation quality across all supported domains.

The model demonstrates substantial advancements in tasks requiring complex, step-by-step reasoning, and its integrated thinking modes allow for adjusting the depth of analysis based on query complexity. With support for 119 languages and superior semantic extraction capabilities, it is especially effective for multilingual document analysis and cross-linguistic tasks.

Qwen3-4B is ideally suited for professional applications that require deep document analysis — such as legal review, scientific research, technical documentation, and literary analysis. The model excels at summarization, multilingual translation, and organizing academic content in educational projects.

Announce Date: 29.04.2025
Parameters: 4B
Context: 131K
Attention Type: Full or Sliding Window Attention
VRAM requirements: 19.9 GB using 4 bits quantization
Developer: Alibaba
Transformers Version: 4.51.0
Ollama Version: 0.6.6
License: Apache 2.0

Use our pre-built public endpoints to test inference and explore Qwen3-4B capabilities.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.