Opus 5 improves coding, reasoning efficiency, and prompt-cache-friendly tool use, and is priced at $5/1M input tokens and $25/1M output, same as Opus 4.8
I do not have to sell my kidney for using it? [embedded post]
Context & Ripple Effects
Opus pricing had already moved sharply lower when Opus 4.5 adopted $5/$25 per million input/output tokens, versus the earlier Opus 4.1 level. Opus 5 keeps that price point while claiming better coding and reasoning efficiency.
The release also follows an Opus 4.6 fast-mode preview that traded substantially higher cost for speed. By contrast, the reported emphasis on prompt-cache-friendly tool use targets the cost and efficiency of repeated, tool-using workloads rather than a premium speed tier.
First-order effects
- Anthropic customers can access Opus 5’s reported coding and reasoning improvements without a listed per-token increase over Opus 4.8: $5 per million input tokens and $25 per million output tokens.
- Prompt-cache-friendly tool use can reduce repeated-input overhead for applications with recurring context, improving the effective cost of Opus-based agent workflows where caching is applicable.
Second-order effects
- Developers evaluating frontier models for coding agents will compare task throughput and cache behavior—not headline token rates alone—because a stable list price can still lower cost per completed task.
- Rival model providers face pressure to pair capability upgrades with clearer efficiency gains or pricing discipline; the prior Opus 4.6 fast mode illustrates that speed-oriented offerings can carry a distinct cost trade-off.
Third-order effects
- If frontier-model upgrades continue to arrive at unchanged list prices, competition will increasingly turn on effective inference cost: how efficiently models reason, reuse context, and complete tool-driven work.
- This reinforces a split between standard tiers optimized for broad production economics and premium capacity or latency options, though the durability of that split depends on providers’ compute costs and demand.
The trend: Frontier AI pricing is shifting from price-per-token competition toward cost per useful, tool-completed task, with caching and inference efficiency becoming key differentiators.