Qwen/Qwen3-30B-A3B-Instruct-2507

Qwen/Qwen3-30B-A3B-Instruct-2507

Efficient Mixture-of-Experts (MoE) Instruction-Finetuned Language Model
2025-07-29
LLM
Model capability: function_call
Input:
$0.1/1M tokens
Output:
$0.4/1M tokens
Bulk order? Contact your manager for exclusive deals

API Overview

Qwen3-30B-A3B-Instruct-2507 is an efficient mixture-of-experts (MoE) instruction-tuned language model released by Alibaba’s Tongyi Lab. Its core positioning is as a high-performance, cost-effective, enterprise-grade general-purpose reasoning powerhouse characterized by “small activation, great capability, and high cost-effectiveness.”

  • Exquisite MoE Architecture: With a total parameter count of approximately 30 billion, it activates only 3 billion parameters (A3B), significantly reducing computational costs and inference latency while maintaining powerful overall capabilities.
  • 128K Ultra-Long Context: Natively supports ultra-long text inputs, making it ideal for scenarios such as long-document summarization, multi-turn complex dialogues, and cross-paragraph information integration.
  • Deeply Optimized Instruction Following: Fine-tuned with high-quality human preference data, it delivers precise and reliable performance in format control, style imitation, and multi-constraint tasks.
  • Well-Balanced Multilingual and Coding Capabilities: It strengthens Chinese context understanding while supporting mainstream languages including English, Japanese, and French, as well as programming languages such as Python and JavaScript.

───────────────────────────────────────────────────────────────────

Core Capabilities

Ultimate Energy Efficiency: Achieves 30-billion-parameter-level language understanding and generation capabilities with resource consumption comparable to that of a 7-billion-parameter dense model, delivering higher output per unit of computing power.

🧠 Stable Task Execution: Demonstrates outstanding performance in benchmarks such as C-Eval, MMLU, and HumanEval, making it suitable for production environments with high requirements for determinism.

🌍 Natural Bilingual Expression in Chinese and English: Produces fluent and natural outputs, seamlessly balancing technical terminology and cultural context, ideal for content creation, intelligent customer service, office automation, and other applications.

🛡️ Enterprise Security and Compliance: Supports private deployment, content filtering, and audit logs, meeting regulatory requirements in industries such as finance, government, and education.

Playground

Log in to explore more features! Click to Log In

API Analytics

API Reference (1)

API DescriptionAPI EndpointRequest MethodStabilityParameter Description
Chat(SiliconFlow)
POST
Stable
View Details

API Pricing

$
ModelDescriptionContextOfficial Price302.AI Price

Qwen/Qwen3-30B-A3B-Instruct-2507

-
256000

Input$0.1 / 1M tokens
Output$0.4 / 1M tokens

Input$0.1/ 1M tokens
Output$0.4/ 1M tokens
Original Price