Amazon engineers are reportedly distilling Anthropic models to build smaller, cheaper versions for internal use, according to a report from The Information. The effort is aimed at reducing costs ahead of a new pricing model that will charge based on tokens processed rather than compute hours starting next year. This approach involves using a smaller model to learn from the outputs of a larger model, allowing Amazon to maintain functionality while minimizing expenses. The move is part of a broader effort to manage costs as the company renegotiates its partnership with Anthropic.
Amazon has certain rights to use Anthropic's models for distillation, similar to Apple's arrangement with Google Gemini. However, Anthropic's Claude models are not available on Amazon's Bedrock cloud platform, which currently supports only Amazon's own Nova models and Meta's Llama models. The company is also reportedly exploring alternatives like OpenAI and its own Nova models. Amazon has invested up to $25 billion more in Anthropic and up to $50 billion in OpenAI this year.
An Amazon spokesperson said the changes from the expanded partnership won't raise costs, while Anthropic highlighted lower prices relative to the performance its models deliver. The Information reported that the effort ties back to a renegotiation of the partnership, which could impact future cost structures. Source: thedecoder
