AWS adds fully managed AI models: Qwen3 and DeepSeek-V3.1

AWS adds fully managed AI models: Qwen3 and DeepSeek-V3.1

Amazon Web Services has expanded its Amazon Bedrock portfolio by adding fully managed open-weight models Qwen3 and DeepSeek-V3.1, signaling a strategic shift toward greater flexibility for enterprise customers. This move allows developers to access models with increased transparency regarding weights, facilitating easier customization and fine-tuning for specific business requirements. By integrating these models, AWS provides a diverse array of options that cater to varying computational needs, from sophisticated coding tasks to complex reasoning scenarios, ensuring businesses can select tools that best align with their unique operational goals. The introduction of these open-weight models highlights significant advantages in cost efficiency and localization, particularly for international markets in Asia and Latin America. Unlike large, proprietary general-purpose models, open-weight variants enable organizations to distill or fine-tune models into smaller, cheaper versions without sacrificing substantial accuracy. This capability empowers startups and established enterprises to build specialized small language models tailored to local languages and niche use cases, avoiding high licensing fees while maintaining data privacy through AWS’s secure, enterprise-grade infrastructure. This development is crucial for the open_data movement as it demonstrates how cloud providers can democratize access to advanced AI capabilities. By offering open-weight models on a managed platform, AWS bridges the gap between transparency and accessibility, allowing developers to inspect, adapt, and improve upon underlying model architectures. This fosters a more collaborative and innovative ecosystem where data and model improvements are shared openly, rather than being siloed within proprietary black boxes, ultimately accelerating the adoption of ethical and customizable AI solutions across industries.

Source: siliconangle.com
Published on 2025-09-19