Amazon Bedrock just added three new frontier models, and none of them cost what their predecessors did.
OpenAI's GPT-6 Sol and GPT-6 Luna landed on Bedrock alongside Anthropic's Claude Opus 5.5, the first release in the Opus 5.5 family. Sol is aimed at the recurring, demanding work of development and operations; Luna is built for narrower tasks run at high volume. Both ship at significantly lower prices than their GPT-5.6 predecessors. Opus 5.5 is tuned for agentic coding and long-running tasks, and Anthropic says it uses fewer tokens than Opus 5 to do comparable work - a token-efficiency claim, not a speed claim.
The interesting part isn't which model is smartest. Bedrock is betting most teams will run a mix of models, picking the cheapest one that clears the bar for a given step instead of defaulting to the flagship every time. That's a shift from the biggest-model-wins framing that has dominated AI marketing toward something closer to ordinary cloud economics: match the resource to the job.
It's also a quiet admission that model providers are now competing on cost per task as much as raw capability, which happens to be the number that actually shows up on your bill.