Amazon Web Services Expands AI Model Portfolio
Amazon Web Services (AWS) has announced the global availability of the open weight Qwen3 models from Alibaba and DeepSeek-V3.1 in Amazon Bedrock. This move significantly enhances AWS’s extensive selection of fully managed open weight models, joining offerings from Meta, Mistral AI, and OpenAI.
These models are now accessible to customers worldwide, underlining AWS’s commitment to providing a diverse range of AI tools that cater to various enterprise needs. Customers can securely deploy these state-of-the-art models with enterprise-grade data protection, encryption, and privacy controls.
Open Weight Models Bring Transparency and Flexibility
Open weight models allow developers to view the internal workings, or “weights,” of the AI systems. This transparency helps customers tailor models to their unique needs, unlocking innovation and improving cost efficiency. AWS ensures that customer inputs and outputs are not shared with model providers and are not used to retrain the base models.
With Amazon Bedrock, customers can also implement Amazon Bedrock Guardrails to detect and mitigate hallucinations and other potential model issues, enhancing the reliability of AI outputs.
Meet the Qwen3 Family from Alibaba
The Qwen3 family features four new open weight models designed for varied applications:
- Qwen3-Coder-480B-A3B-Instruct and Qwen3-Coder-30B-A3B: Suited for software engineering, these models support multi-language code generation and can autonomously interact with APIs and tools.
- Qwen3-235B-A22B-Instruct-2507: A general-purpose model optimized for reasoning, math, and coding with a balance of performance and efficiency.
- Qwen3-32B (Dense): Ideal for resource-constrained environments, offering consistent and predictable performance.
These models support multilingual capabilities, excelling in both English and Chinese, making them ideal for global operations and cross-cultural content generation.
Introducing DeepSeek-V3.1
DeepSeek-V3.1 is the latest in the DeepSeek series, boasting 685 billion parameters and a “mixture of experts” (MoE) architecture. This design activates only the necessary parts of the model for each task, making it highly efficient while preserving top-tier performance.
The model offers two distinct modes: a “thinking mode” for complex problem-solving and a “quick response mode” for simpler queries. This flexibility allows customers to toggle between speed and depth depending on use case.
DeepSeek-V3.1 is particularly effective in software development, mathematical reasoning, data analysis, and building autonomous agents, making it a powerful tool for developers and enterprises alike.
Real-World Applications and Customer Success
Organizations like CoinMarketCap are already leveraging DeepSeek models in Amazon Bedrock to process and gain insights from vast amounts of cryptocurrency data. These capabilities are helping companies enhance operations and deliver more sophisticated services.
Open weight models also democratize AI by removing the need for specialized teams, enabling more industries to adopt advanced AI solutions. Whether for healthcare, finance, manufacturing, or research, these models can power automated workflows, in-depth analysis, and innovation at scale.
Why Open Weight Models Matter
Unlike proprietary models, open weight models serve as a transparent foundation—like a soup base that can be customized with additional “ingredients” to meet specific needs. This flexibility fosters innovation by allowing developers to build on existing work instead of starting from scratch.
According to Luis Wang, Director of Amazon Bedrock, “Open weight models represent a vital frontier in AI innovation. We don’t think there’s going to be one model that’s the best for every use case. Open models offer the flexibility to experiment and customize.”
Technical Capabilities and Insights
The Qwen3-Coder-30B-A3B-Instruct and Qwen3-235B-A22B-Instruct-2507 models can process up to 262,000 tokens simultaneously—equivalent to around 200,000 words or two entire novels in a single session.
DeepSeek-V3.1’s 685 billion parameters allow it to understand and solve complex problems with remarkable precision while reducing compute costs by activating only relevant parameters for each task.
Artificial Analysis notes that DeepSeek-V3.1 shows marked improvements over previous generations, especially in agentic reasoning and problem-solving tasks.
Expanding Global Access
Following the integration of OpenAI’s models, AWS is now rolling out these new open weight models across every AWS Region. These include locations such as US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo, Mumbai), Europe (Ireland, London, Milan, Stockholm), and South America (São Paulo).
By making these powerful models widely accessible, AWS is enabling businesses without deep AI expertise to implement advanced solutions and accelerate innovation across sectors.
Looking Ahead
AWS’s ongoing investment in a diverse AI ecosystem reflects its mission to unlock the full potential of generative AI. With more than 135 AI and machine learning training courses available, AWS is also helping professionals at all levels upskill and prepare for the future of AI.
To explore these new models, users can visit the Amazon Bedrock console and begin integrating them into their solutions today.
This article is inspired by content from Original Source. It has been rephrased for originality. Images are credited to the original source.