Anthropic has fundamentally restructured its Claude lineup, introducing three distinct models at different price points and capability levels that require different decisions for different workflows. Claude Fable 5 launched June 9, 2026, was pulled by the US government three days later due to cybersecurity concerns, and returned July 1 with tighter safety guardrails. Claude Opus 4.8 arrived in late May as the workhorse for serious knowledge work. Claude Sonnet 5 launched June 30 as the affordable default. The wrong choice between them can cost you 5 to 10 times what you should be spending, or leave you paying premium prices for tasks a cheaper model handles equally well.

What Separates Claude Fable 5, Opus 4.8, and Sonnet 5?

Before June 2026, Anthropic's model lineup was straightforward: Opus for maximum capability, Sonnet for balance, and Haiku for speed and cost. That structure flipped when Anthropic introduced a fourth tier called the Mythos class, sitting above Opus. Claude Fable 5 is the publicly available Mythos-class model, which means it is not simply a bigger Opus; it is a different tier entirely.

The three models now available represent fundamentally different use cases and economics:

• Claude Fable 5: Designed for days-long autonomous agent tasks with a 1 million token context window, up to 128,000 output tokens per request, and costs $10 per million input tokens and $50 per million output tokens. It includes mandatory 30-day data retention with no zero-data-retention option available.
• Claude Opus 4.8: Built for serious knowledge work and complex coding with a 200,000 token context window, up to 64,000 output tokens per request, and costs $5 per million input tokens and $25 per million output tokens. It offers zero data retention options unlike Fable 5.
• Claude Sonnet 5: Released June 30 as the new default across consumer and developer plans, positioned as a meaningful step up in agentic performance with stronger tool use and better multi-step coherence. It delivers most of Opus's capability at a fraction of the cost, making it best for high-volume workflows and teams watching their token budget.

Why Did the US Government Pull Claude Fable 5?

On June 12, 2026, the Trump administration ordered Anthropic to suspend both Fable 5 and Mythos 5 due to concerns about cybersecurity misuse potential. For nearly three weeks, Fable 5 was offline. When the model returned globally on July 1, it came with reinforced safety classifiers that now catch and reroute more edge cases than the original launch version. The re-released Fable 5 is slightly more restrictive than its initial release, but it still represents Anthropic's most capable widely available model.

This government intervention signals something important about how frontier AI capabilities are being regulated. Anthropic believes Mythos-class capabilities are commercially valuable enough to charge premium prices, and the company is building runway for even higher tiers in future generations. The safety guardrails added after the government intervention reflect a conservative tuning approach; Anthropic estimates that safety classifiers catch benign requests in less than 5 percent of sessions on the original release, though the July 1 re-release is stricter.

When Should You Actually Use Claude Fable 5?

Claude Fable 5 excels at specific, high-value use cases where its frontier capabilities justify the 2x pricing compared to Opus 4.8. Boris Cherny, who built Claude Code, described Fable 5 as "the first model I have used that was so methodical and precise, taking measurements and adding logs, then verifying that it truly fixed the issue before declaring victory".

"The first model I have used that was so methodical and precise, taking measurements and adding logs, then verifying that it truly fixed the issue before declaring victory," said Boris Cherny, who built Claude Code.

Boris Cherny, Claude Code Developer

Fable 5 is built for "days-long, complex, and asynchronous tasks previous models couldn't sustain." When deployed inside agent harnesses like Claude Code or Claude Managed Agents, Fable 5 plans across stages, delegates to sub-agents, writes its own tests, and verifies its work using vision before declaring completion. Anthropic's own benchmarks show Fable 5 outperforming Opus 4.8 on complex coding tasks, especially those requiring multi-hour autonomous execution.

However, Fable 5 comes with real constraints. Usage limits apply through July 7, 2026; Claude Pro, Max, Team, and Enterprise subscribers can use up to 50 percent of their weekly usage limit on Fable 5, after which additional access requires purchasing credits. Data retention is not optional; both Fable 5 and Mythos 5 are Anthropic's "Covered Models," meaning mandatory 30-day prompt and output retention with no zero-data-retention option available. For regulated industries or clients with strict data governance requirements, this alone can be a dealbreaker.

How to Choose the Right Claude Model for Your Workflow

• Choose Fable 5 if: You run long-horizon coding tasks, days-long autonomous agents, or need the absolute frontier of AI capability. The additional 25 to 30 percent task completion speed should pay for the 2x pricing in your specific workflow.
• Choose Opus 4.8 if: You do serious knowledge work, complex coding, or enterprise-grade analysis. It is the workhorse: reliable, powerful, and available without Fable's usage restrictions or mandatory data retention requirements.
• Choose Sonnet 5 if: You want most of Opus's capability at a fraction of the cost. It is best for high-volume workflows, agentic tasks that do not require frontier reasoning, and teams watching their token budget. Most workflows should default to Sonnet 5.

Claude on Azure: How Enterprise Procurement Just Changed

Beyond the model lineup restructuring, Anthropic has cleared a major institutional barrier to enterprise Claude deployment. Claude is now generally available in Microsoft Foundry, hosted natively on Azure, billed through Azure, and governed by the identity and access policies enterprises already operate. The announcement, which came in two steps on June 29 and July 1, removes the need to sign a separate Anthropic commercial contract and manage a second vendor relationship.

This matters because most enterprise AI projects do not stall because of model quality. They stall because of "everything around the model: procurement, governance, networking, and data," according to Azure Product Lead Steve Sweetman. Before this week, an enterprise developer wanting to deploy Claude through Azure could reach Anthropic's API, but the billing, identity management, and security review sat outside Azure's governance envelope. A central IT team following standard vendor approval processes had to treat Anthropic as a distinct vendor with its own security assessment, data processor agreement, and invoice. That friction kept many organizations in a pilot-indefinitely state.

The June 29 and July 1 releases collapse that friction. Teams authenticate with Microsoft Entra ID, apply Azure role-based access controls, and receive a single consolidated Azure invoice with per-model detail. Claude usage is billed in Claude Consumption Units as a single line on the Azure Marketplace bill. For eligible customers with a Microsoft Enterprise Agreement, that usage draws down existing Microsoft Azure Consumption Commitment credits, converting a previously off-ledger AI tool into a spend category their finance teams already manage.

Claude Opus 4.8 and Claude Haiku 4.5 reached general availability in Foundry on June 29, with Claude Sonnet 5 following on July 1. All three Claude models run in Foundry on NVIDIA Blackwell Ultra hardware with MACC billing and Entra ID authentication.

The structural problem this solves is shadow AI. When central IT cannot approve an external AI tool quickly enough because the security review, data processor agreement, and vendor vetting take months, product teams and developers use it anyway through personal accounts or unofficial channels. The result is a governance gap: usage the organization cannot audit, data it cannot account for, and cost it cannot attribute. Claude's integration into Azure's existing governance stack directly addresses this pattern.