Opus 5.5: Cheaper, But Thinks Longer

🔥 What happened Anthropic dropped Opus 5.5: input and output tokens are 20% cheaper than Opus 5, cache reads a full 60% cheaper. The model now always thinks before answering. 💡 Why it matters Cache reads cost a twentieth of input price instead of a tenth. In a typical Claude Code session with 90% cache hits, input drops from $1.00 to $0.40. But one retry can wipe out the entire saving. ⚡ Our take Cheaper tokens are nice, but if you ignore turn count and cache hits, you still pay more. The real message: prompt discipline beats the price sheet.