Alibaba Cloud Qwen3 Models Now Available on Amazon Bedrock for Smarter AI Workflows

Overview

Amazon Bedrock has added four Qwen3 models from Alibaba Cloud, offering fully managed, serverless access to sophisticated AI capabilities. Introduced on September 18, 2025, the portfolio consists of Qwen3-Coder-480B-A35B-Instruct, Qwen3-Coder-30B-A3B-Instruct, Qwen3-235B-A22B-Instruct-2507, and Qwen3-32B Dense, both with Mixture-of-Experts (MoE) and dense architectures. The models support repository-scale code analysis, agent-like workflow automation, and cost-effective AI applications without infrastructure management.

Introduction

The fusion of Qwen models within Amazon Bedrock makes top-tier AI accessible to the world’s businesses. Alibaba’s Qwen3 lineup brings leading-edge coding, multilingual processing, and agentic capabilities into AWS’s managed platform.

Powerful apps for code generation, autonomous agents, and business automation are now developed by organizations with full control over their data.

Model Portfolio

Qwen3-Coder-480B

The leader Qwen3-Coder-480B-A35B-Instruct has a total of 480 billion parameters with 35 billion active via MoE architecture. Designed for agentic coding and advanced software engineering, it delivers state-of-the-art performance in agentic coding, browser usage, and tool orchestration tasks. It natively supports 256K context length with extensibility to 1M tokens, enabling end-to-end repository-scale analysis and multistep workflow automation. It excels in code generation, debugging, refactoring, and intricate algorithm implementation for Python, JavaScript, Java, C++, Go, and Rust.

Qwen3-Coder-30B

The Qwen3-Coder-30B-A3B-Instruct offers effective coding functionality with a total of 30 billion parameters and 3 billion active parameters. Designed for generating, analyzing, and debugging code, it offers powerful performance with reduced computational demands. Each Coder model includes sophisticated function calling and tool utilization, making it well-suited for creating self-coding agents.

Qwen3-235B General Model

Qwen3-235B-A22B-Instruct-2507 harmonizes ability with effectiveness via 235 billion total parameters and 22 billion active. It performs well on CodeForces ELO Rating, BFCL, and LiveCodeBench v5 benchmarks, proving competence in competitive coding and real-world programming. With support for more than 29 languages with 128K token context, it provides powerful multilingual applications.

Qwen3-32B Dense

The Qwen3-32B Dense features standard dense architecture with 32 billion parameters and is ideal for resource-limited environments and edge installations. It features consistent performance and easy deployment, where constant latency is paramount.

Key Capabilities

Agentic Coding Excellence

Qwen3 models manage entire codebases with rich contextual comprehension, parsing project layouts, recognizing patterns, and producing uniform implementations for giant systems. Unparalleled debugging functionality automatically detects flaws and delivers detailed resolutions. Leading-edge tool utilization benchmarks enable advanced autonomous agents to manage multiple development tools and APIs.

Extended Context Processing

Native 256K token support for up to 1M tokens enables processing large documentation, extensive codebases, and multi-file repositories within a single context window. Developers utilize entire project histories, comprehensive API specifications, and extensive documentation without fragmentation. Conversations extended save project state during development sessions.

Multilingual Support

More than 29 languages support delivers native-quality processing for worldwide teams. Features include multilingual code generation, documentation, natural language programming instructions, and localization with technical accuracy. Solid Chinese and English performance supports international organizations.

Enterprise Security

Business-class features provide data protection and compliance. Customer data never gets used to train models, and AWS will not share input/output data with providers. Organizations maintain complete control with audit trails, encryption, and access control. Amazon Bedrock Guardrails stop hallucinations for production reliability.

Performance Benchmarks

Qwen3-Coder scores 9.25/10 on clean markdown tasks, 89.3% pass@1 on HumanEval, and 78.2% on MBPP. Qwen3-235B is top-ranked in CodeForces ELO Rating in competitive programming. Independent testing reveals performance comparable to or higher than that of Claude Sonnet 4 and GPT-4 on complex, multi-turn, agentic tasks. High BFCL scores indicate consistent use of browser-based tools and the ability to engage in multistep reasoning.

Deployment and Pricing

Regional Availability

Qwen3-32B and Qwen3-Coder-30B: US East (N. Virginia), US West (Oregon), Asia Pacific (Mumbai, Tokyo), Europe (Ireland, London, Milan, Stockholm), South America (São Paulo). Qwen3-235B: US West (Oregon), Asia Pacific (Mumbai, Tokyo), Europe (London, Milan, Stockholm). Qwen3-Coder-480B: US West (Oregon), Asia Pacific (Mumbai, Tokyo), Europe (London, Stockholm).

Cost Efficiency

Qwen3-Coder-480B: $1.50 per million input tokens, $7.50 per million output tokens (average $3.00/million at 3:1 ratio). The MoE architecture saves money by enabling only the required parameters to be turned on. Dynamic pricing accommodates provisioned and on-demand throughput.

Use Cases

Library Development

Computer-aided code generation, smart review, technical debt reduction, modernization of legacy applications, generation of API specifications, and consistency of microservices. Integration detects weaknesses, recommends optimization, and generates automatic testing. Analysis at the repository scale enables refactoring and knowledge sharing.

Business Automation

Highly advanced autonomous agents manage processes, coordinate tools/APIs, and run multistep workflows. Intelligent automation learns to adapt to circumstances, manage exceptions, and learn operating patterns.

Conclusion

Qwen models in Amazon Bedrock enable accessible, enterprise-grade AI for software development and automation. Qwen3, integrated with AWS infrastructure, provides high-impact tools for coding agents, multilingual apps, and autonomous workflows without infrastructure overhead. The portfolio, which ranges from a 480B-parameter Coder to a fast 32B dense variant, caters to diverse needs. With competitive pricing, enterprise security, wide regional availability, and leading-edge performance, Qwen models help organizations drive AI adoption at scale.

Drop a query if you have any questions regarding Qwen models and we will get back to you quickly.

About CloudThat