Anthropic launches Claude Opus 4.8 – UK pricing and details

Anthropic has released Claude Opus 4.8, the latest version of its flagship AI model.

The group said that the upgrade builds on Opus 4.7 with measurable gains across benchmarks and is positioned as a more effective collaborator for complex tasks.

It is available immediately at the same price as its predecessor but will cost more tokens to use.

Key improvements and new features

Opus 4.8 launches alongside several practical enhancements:

• Effort Control on claude.ai and Cowork: Users can now adjust how much “thinking” Claude does per task. Higher effort settings deliver deeper reasoning at the cost of more tokens and time, while lower settings provide faster responses and slower rate-limit consumption.
• Dynamic Workflows in Claude Code (research preview): The model can plan large projects, spin up hundreds of parallel sub-agents, verify outputs, and handle codebase-scale tasks end-to-end. This is particularly powerful with Opus 4.8’s extended agent runtime.
• Faster Fast Mode: The high-speed variant of Opus 4.8 now runs at 2.5× the speed of previous models and is three times cheaper than fast mode on earlier Opus versions.

Capabilities

Anthropic saidOpus 4.8 shows solid gains in coding, agentic skills, reasoning, and knowledge-work tasks.

Early testers report the model feels more reliable and sharper in judgment, especially on agentic workflows. and a notable strength is improved honesty.

Opus 4.8 is significantly less prone to overconfident but unsupported claims. Evaluations show it is around four times less likely than Opus 4.7 to let flaws in generated code pass without comment.

Alignment testing indicates Opus 4.8 reaches new highs in prosocial traits – supporting user autonomy and acting in the user’s best interest – while showing lower rates of misaligned behaviour than Opus 4.7, comparable to Anthropic’s best-aligned preview models.

Also launching today

• Dynamic workflows (Claude Code Enterprise, Team, and Max plans): Enable massive parallel agent runs with built-in verification.
• Messages API update: System instructions can now be updated inside the messages array without breaking prompt cache.
• Increased rate limits in Claude Code to support higher-effort usage.

Opus 4.8 defaults to a “high effort” setting that Anthropic considers the best balance of quality and experience. Users can push to “extra” or “max” for tougher tasks.

Pricing (unchanged for regular usage)

• Standard mode: $5 per million input tokens / $25 per million output tokens
• Fast mode: $10 per million input tokens / $50 per million output tokens

UK users pay the same USD rates via the Claude API and claude.ai (converted to local currency at checkout). Enterprise and Team plans include the new features.

Plans are priced in USD (shown on the site), with billing converted to GBP + VAT where applicable. Current rates:

• Free: $0 – Limited usage
• Pro: $20/month (billed monthly) or $17/month ($200 billed annually upfront) – Includes Claude Code, Cowork, higher limits, and access to Opus 4.8 on higher effort settings. claude.ai
• Max: From $100/month (5× Pro usage) up to $200/month (20× Pro usage) – Recommended for heavy Claude Code / dynamic workflows use.

What’s next?

Anthropic describes Opus 4.8 as a “modest but tangible” step forward and said it is already working on more cost-effective models with similar capabilities.

The company also plans to release higher-intelligence “Mythos-class” models in the coming weeks, starting with organisations already using Claude Mythos Preview for cybersecurity under Project Glasswing.

Stronger cyber safeguards are being finalised before wider release.

Full technical details, benchmarks, and the complete alignment assessment are published in the Claude Opus 4.8 System Card.

Now read: AI Engineer job postings jump 81% in the UK as firms scramble for talent