What’s actually changing
Claude Opus 4.8 launched 2026-05-28 at the same base pricing as its predecessor: $5 per million input tokens and $25 per million output tokens, according to third-party pricing-tracking coverage (finout.io), not confirmed against Anthropic’s own pricing page directly. On its own, that reads as no change at all.
Fast Mode tells a different story. The same source reports Fast Mode’s price fell from $30 per million input tokens and $150 per million output tokens under Opus 4.7, to $10/$50 under Opus 4.8, a figure that matches Anthropic’s own Claude pricing page. That is a real reduction in a mode specifically built for latency-sensitive workloads that pay a premium to get responses faster.
Structural Comparison Matrix
| Operational Aspect | Opus 4.7 | Opus 4.8 |
|---|---|---|
| Base price (input/output per MTok) | $5 / $25 | $5 / $25 (unchanged) |
| Fast Mode price (input/output per MTok) | $30 / $150 | $10 / $50 |
| Tokenizer | Introduced in this version | Inherited unchanged from Opus 4.7 |
Fix it: re-run the Fast Mode math
Anyone who evaluated Fast Mode against Opus 4.7’s $30/$150 pricing and decided the latency benefit wasn’t worth the cost has a real reason to revisit that decision under Opus 4.8. A workload that pays for Fast Mode specifically to cut response time now does so at roughly a third of the previous rate, which changes the break-even point against standard mode meaningfully, not marginally.
Base pricing staying flat is worth noting for a different reason: it means the price cut is isolated to Fast Mode specifically, not a general Opus 4.8 discount. Don’t assume standard-mode costs moved just because Fast Mode did.
The LLM Pricing Calculator has both Opus 4.7 and 4.8, standard and Fast Mode, as selectable options — a faster way to re-run this exact break-even math against your own token counts than doing it by hand.
Same price doesn't mean same cost per prompt
Opus 4.8 inherited Opus 4.7’s tokenizer unchanged, and that tokenizer is reported to count noticeably more tokens for the same text than older models. See Claude’s New Tokenizer Counts Up to 35% More before assuming your cost per prompt is unchanged.
Confirmed version
Both the base-price figure and the Fast Mode figures are paraphrased from finout.io’s Claude Opus 4.8 pricing breakdown, tied to the model’s 2026-05-28 launch, not confirmed against Anthropic’s own pricing page directly. Browse more coverage in the AI Productivity archive, or start from The 2026 LLM Token & Pricing Reset hub.
