Qwen3-30B-A3B-Instruct-2507

Qwen3-30B-A3B-Instruct-2507 is an interactive language model specifically optimized for dialogue and instruction-following tasks. This updated version of Qwen3-30B-A3B employs the same Mixture-of-Experts (MoE) architecture with 30.5 billion total parameters and 3.3 billion active parameters, but omits explicit reasoning steps and is tuned for instantaneous response generation. Architectural enhancements include native support for an extended context length of up to 262,144 tokens, with a recommended output length of 16,384 tokens per generation.

The model achieves outstanding performance of 90.0 on the ZebraLogic benchmark, significantly surpassing both DeepSeek-V3 (83.4) and GPT-4o (52.6). This demonstrates the model’s strong ability to produce logically coherent and well-justified responses without relying on step-by-step reasoning. In creative tasks, the model excels with scores of 86.0 on Creative Writing v3 and 85.5 on WritingBench, outperforming GPT-4o and Gemini-2.5-Flash respectively, making it an unmatched tool for generating high-quality creative content. Additionally, the model performs exceptionally well in programming tasks, achieving 83.8 on MultiPL-E, showcasing its capability to generate high-quality code across various programming languages.

These features and capabilities make Qwen3-30B-A3B-Instruct-2507 the ideal choice for applications requiring lightning-fast responses and natural user interaction. It is perfectly suited for chatbots and applications focused on rapidly generating high-quality textual content.

Announce Date: 29.07.2025
Parameters: 30.5B
Experts: 128
Activated: 3.3B
Context: 263K
Attention Type: Full or Sliding Window Attention
VRAM requirements: 37.1 GB using 4 bits quantization
Developer: Alibaba
License: Apache 2.0

Use our pre-built public endpoints to test inference and explore Qwen3-30B-A3B-Instruct-2507 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.