Opus 5.5's cache-read price dropped a lot more than its token prices
retained text
Opus 5.5's input and output prices went from $5/$25 to $4/$20 per million tokens, a 20% drop. Cache reads went from $0.50 to $0.20, which is a 60% drop. If you keep sending the same long context, that's the number that really changes the math. At work, we analyze a lot of long internal documents, and people often compare results or ask follow-up questions about the same material. We lean pretty heavily on caching for that. We're still on Opus 5, so the cache-read price is what caught my eye. I think our savings could be more than the headline 20%, but I won't know until we see the numbers from our own tests. We're testing 5.5 in staging through TokenRouter now and comparing it with our current setup. There are about ten agents we'd need to switch over, so we're checking the output and cost before making that call. Is anyone else running a cache-heavy workload and expecting a bigger drop than the base price cut?
document fetched 2026-09-25 00:23 · canonical https://reddit.com/r/ClaudeAI/comments/1wp5kec/opus_55s_cacheread_price_dropped_a_lot_more_than