List price did not move. Anthropic's 28 May 2026 post puts Claude Opus 4.8 at $5 per million input tokens and $25 per million output tokens, same as Opus 4.7. Fast mode is $10/$50. Benchmarks and honesty claims on that post are Anthropic's.

The news

The Anthropic post "Introducing Claude Opus 4.8," dated 28 May 2026, states Opus 4.8 "builds on Opus 4.7 with improvements across benchmarks, and is a more effective collaborator. It's available today for the same price."

Availability: "Claude Opus 4.8 is available everywhere today." Regular usage: "$5 per million input tokens and $25 per million output tokens." Fast mode: "$10 per million input tokens and $50 per million output tokens," where the model "can work at 2.5 times the speed" and fast mode "is now three times cheaper than it was for previous models." Developers are told to use the Claude API model id claude-opus-4-8.

The post also launches: dynamic workflows in research preview in Claude Code for Enterprise, Team, and Max plans, described as planning work then running "hundreds of parallel subagents in a single session"; an effort control on claude.ai and Cowork on all plans; and Messages API support for system entries inside the messages array. Opus 4.8 "defaults to high effort." Anthropic recommends "extra" (xhigh in Claude Code) or "max" for difficult and long-running jobs, and says it increased Claude Code rate limits for higher effort.

An AWS machine-learning blog dated 28 May 2026 says Opus 4.8 is available on Amazon Bedrock and Claude Platform on AWS, naming Bedrock regions including US East (N. Virginia), Asia Pacific (Tokyo), Europe (Ireland), and Europe (Stockholm), with a pointer to Bedrock docs for the full list.

The Claude Opus 4.8 System Card, dated 28 May 2026 (changelog entries 3 June and 17 June 2026), reports Responsible Scaling Policy evaluations. Anthropic's overall conclusion in the executive summary: Opus 4.8 "does not advance the capability frontier beyond our most capable model (Claude Mythos Preview), and that catastrophic risks from the deployment of this model remain low given our current mitigations." On agentic safety, the summary says Opus 4.8 was "somewhat less robust than Opus 4.7 in several agentic contexts (such as vulnerability to prompt injection attacks)" without safeguards, and that "the application of our safeguards closes the gap between the models in practice."

Who is bound

Anthropic shipped the model, the blog, and the system card. AWS shipped Bedrock and Claude Platform on AWS listings. Buyers on Claude API, claude.ai, Claude Code, Bedrock, and other named clouds consume the new model id and the new fast-mode prices. Enterprise, Team, and Max Claude Code seats are the bound set for dynamic workflows in research preview.

What's new

Fast-mode price did move. Effort control is now a user-visible lever on claude.ai and Cowork. Dynamic workflows are a Claude Code research-preview path for large parallel subagent jobs, not a default for every seat. The system card is the document that constrains safer/smarter marketing: RSP says not past Mythos Preview. Prompt-injection tests vs 4.7 are not a clean win without Anthropic's safeguards.

What it does not settle

Date: 28 May 2026. Price: $5/$25 standard; $10/$50 fast mode. Dynamic workflows: research preview, Enterprise/Team/Max. Region: Anthropic says "everywhere"; AWS names a subset of Bedrock regions. Hardware: none. Data: vendor benches and customer quotes on the launch post; system-card cyber and alignment numbers are Anthropic's. License: proprietary. Mythos-class general release timing in the launch post is "coming weeks." That is a plan, not a ship date.

What to do now

API owners should add claude-opus-4-8 beside 4.7, pin effort, and measure token spend at default high effort before calling the upgrade the same bill. Security should read the system-card agentic-safety note on prompt injection and keep existing injection tests.

Whether Opus 4.8 beats Opus 4.7 on your coding suite is UNKNOWN.