EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
TL;DR
A bug in Amazon Bedrock's Codex GPT-5.6 Sol leads to excessive cache-write costs due to the lack of explicit cache controls, resulting in charges up to 10 times higher than expected.
✦ Why It Matters
Review your Amazon Bedrock usage to identify and mitigate unexpected cache-write costs immediately.
Key Takeaways
How It Works
The Codex CLI generates cache-write tokens based on user requests. When explicit cache controls are not available, every unique request can lead to unnecessary cache writes, inflating costs significantly.
Related