Anthropic Cuts Claude Fast Mode Costs by Two-Thirds

Anthropic's latest model just got significantly cheaper to run. The company launched Claude Opus 4.8 on May 28, 2026, and the headline number is hard to ignore: fast mode now costs one-third what it did with the previous version, Opus 4.7.

That is a 67% price drop. For marketing and communications teams that run AI at any meaningful scale, the economics just shifted.

What Changed in the Pricing

Regular pricing for Opus 4.8 stays the same as Opus 4.7: US$5 per million input tokens and US$25 per million output tokens. The big change is in fast mode, which drops from US$30/US$150 to US$10/US$50 per million tokens.

Fast mode also runs 2.5 times faster than standard mode. So you get dramatically lower cost and dramatically higher speed at the same time. That combination matters if your team uses AI for anything time-sensitive, like real-time content generation, live campaign monitoring, or rapid response drafting.

For most enterprise teams, fast mode is where the real-world cost equation lives. The cut makes running frontier-level AI at volume materially more accessible.

More Than a Pricing Story

The pricing change sits alongside a set of capability upgrades worth understanding.

47% of B2B Firms Cut Marketing Jobs via AI—Quietly

Nearly half of B2B SaaS companies have cut marketing roles due to AI. Wynter's report reveals the trend bypasses senior leaders while eliminating junior pathways, with lasting consequences for talent pipelines.

Mission MediaMission Media

Anthropic introduced dynamic workflows in Claude Code as a research preview. The feature lets the model plan complex tasks, run hundreds of parallel subagents in a single session, and verify the outputs before returning them to the user. Anthropic designed this for large software jobs like codebase-wide migrations. But the underlying mechanics, running many parallel work streams simultaneously and checking the results automatically, translate to any high-volume content or workflow automation task.

Dynamic workflows are currently available for Enterprise, Team, and Max plan subscribers.

Opus 4.8 also ships with a new effort control in claude.ai and Cowork. Users can dial up or down how much effort the model puts into a response. Higher settings deliver deeper reasoning. Lower settings produce faster replies and use rate limits more slowly. The default setting is high-effort, which Anthropic says gives the best balance of quality and speed for most tasks.

The Honesty Upgrade That Often Gets Missed

There is a capability change buried in the Opus 4.8 release that deserves more attention from communications leaders than it typically receives.

Anthropic says the model is significantly less likely to overstate progress or draw conclusions without enough evidence. Internal evaluations found Opus 4.8 was about four times less likely than Opus 4.7 to let flaws in its own code pass without comment. The alignment team also found lower rates of deception and cooperation with misuse, with performance comparable to Claude Mythos Preview on those measures.

For anyone using AI in media relations, crisis communications, or executive messaging, the direction that improvement takes is meaningful. A model less prone to making unsupported claims is a more defensible tool for drafting content that will face scrutiny.

What This Means for Budget Planning

The 41% of Asia Pacific organizations that cite budget as their top barrier to AI adoption now have a clearer path forward. KPMG's Asia Pacific AI Adoption Report found that APAC firms plan to invest an average of US$245 million in AI over the next 12 months, well above the global average of US$186 million. Reducing fast mode costs by two-thirds removes one of the most tangible obstacles for mid-size teams that want frontier capability without frontier infrastructure spend.

Deepfakes Are a Distraction. The Real Threat Is Selectively True Narratives.

Deepfakes aren't the real problem. CMOs are ignoring a more dangerous threat: selectively true narratives weaponized to undermine reputation.

Mission MediaMission Media

The strategic question PwC puts to marketing leaders is the right one: are you deploying AI to cut costs, or to multiply what your team can produce? With fast mode now at US$10 per million input tokens, the answer no longer needs to be either-or.