Deploy DeepSeek-R1 Distilled Llama models in Amazon Bedrock

AWS has announced the ability to deploy DeepSeek-R1 Distilled Llama models in Amazon Bedrock, enabling organizations to use efficient, open foundation models with custom model import capabilities.

• DeepSeek AI offers distilled models in 8B and 70B parameter sizes, derived from Llama architectures
• Models can be imported through Amazon Bedrock Custom Model Import and deployed serverlessly
• Distilled models offer performance trade-offs between computational efficiency and reasoning capabilities
• Pricing is based on model copy activity, with minimal charges for storage and inference
• Distilled models maintain 59-92% of original model performance with significantly reduced resources

The solution provides organizations a flexible, cost-effective way to leverage generative AI models within AWS's secure infrastructure.