1 Answer
- Newest
- Most votes
- Most comments
2
As to my understanding, the only option here, if you are building an agent, is to continue to rely on the models.dev JSON or the LiteLLM model cost map. These community projects are currently the only way to get 'real-time' updates on these beta flags without manually checking AWS RSS feeds.
Note: Even if the AWS docs don't show the price, keep in mind that using the 1M context via Bedrock often triggers a long-context premium rate once you pass the 200k token mark - something LiteLLM usually tracks correctly in their pricing.json.
Relevant content
- asked 8 months ago
- asked 2 years ago
Yeah, that's what I thought too, but the pricing page specifically uses 1M context windows (if you look at the anthropic tab https://aws.amazon.com/bedrock/pricing/) - i dont see any mentions of 200k which makes it hard to trust the open source trackers unless they are actively maintained by the people that have the most recent knowledge of the fact. I understand this was the case with 4 and 4.5 rollouts, but things change so quickly and now anthropic removed the premium pricing from 4.6 Sonnet and Opus which I would assume is the case for the provider, otherwise who's earning that multiplier? AWS? Just for the sake of it?