Claude Opus 5 vs Fable 5: Full Comparison

Anthropic released Claude Opus 5 on July 24, 2026. It matches or beats Fable 5 on most published benchmarks at exactly half the price.

The choice between them is no longer about capability ceiling. It is about which workload each model is actually designed for.

This applies directly to Claude Code users: Opus 5 is now the default model for Claude Max subscriptions.

Key Takeaways

• Opus 5 leads or ties Fable 5 on most benchmarks: Agentic coding, knowledge work, computer use, and novel problem solving.
• Fable 5’s remaining advantage is long-horizon autonomy: Multi-day autonomous runs where the model plans, verifies, and self-corrects over extended sessions.
• Pricing: Opus 5 at $5/$25, Fable 5 at $10/$50 per million tokens: A generation-level jump with no price increase over Opus 4.8.
• Model ID for Opus 5: claude-opus-5
• ARC-AGI-3 result: Opus 5 scored 30.2%, roughly triple the previous best result on this benchmark.
• For most production workloads: Opus 5 is the right default. Switch to Fable 5 when the work spans multiple days of uninterrupted autonomous execution.

What Is Claude Opus 5?

Opus 5 is Anthropic’s new everyday flagship model, released July 24, 2026. It replaces Claude Opus 4.8 as the top-tier model available to Claude Pro and Max subscribers.

Anthropic describes it as “designed to be used every day,” which distinguishes it from Fable 5, which Anthropic positions for “your most ambitious work.”

For the first time in the Anthropic model lineup, the everyday model beats the flagship model on most published benchmarks. The distinction between Opus 5 and Fable 5 is no longer primarily capability; it is workload type.

Pricing comparison:

Opus 5 does not require the 30-day data retention policy that Fable 5 users must opt into. For teams with data handling requirements, this is a meaningful practical difference.

Benchmark Comparison

Treat published benchmarks as directional. Effort level, tool access, and agent harness configuration move real production results more than leaderboard deltas. That said, the pattern across independent evaluations is consistent.

Anthropic intentionally weakened Opus 5’s offensive cybersecurity capability while improving general coding and reasoning. The model shows approximately 85% fewer safety classifier interventions than Fable 5 in production, which matters for coding workflows where frequent classifier interruptions slow work down.

What Fable 5 Was, and What Happened to It

Fable 5 launched as Anthropic’s highest-capability publicly available model before Opus 5. It is still available and still the right choice for specific workloads.

One notable context: Fable 5 was briefly taken offline from June 12 to July 1, 2026, following a US government export control directive citing national security. Anthropic restored global access after approximately two weeks of negotiations and added a new safety classifier to both Fable 5 and Mythos 5 as part of that process.

Fable 5 remains Anthropic’s recommended model for:

• Multi-day autonomous agent runs where the model must plan, verify, and self-correct without human steering
• The highest-complexity architectural decisions where choosing the wrong direction is the primary cost
• Workloads where the 0.5% quality edge on CursorBench justifies the 2x price

When to Use Opus 5

Opus 5 is the right default for the vast majority of production workloads in 2026.

Use Opus 5 for:

• Feature implementation and debugging in Claude Code
• Browser automation and computer use tasks
• Document work, structured research, and knowledge work
• Agentic workflows that run for hours, not days
• Any workload where cost per task matters (Opus 5 wins most cost-versus-performance comparisons)
• Coding agent pipelines where safety classifier interruptions were previously a friction point

If Opus 5 completes a task cleanly, its price advantage over Fable 5 is hard to ignore. If it chooses the wrong architecture and requires three restarts, one careful Fable 5 run may end up cheaper. The right metric is cost per finished, approved work, not cost per token.

Opus 5 in Claude Code:

Opus 5 is now the default model for Claude Code on Claude Max subscriptions, and the best model available to Claude Pro subscribers. For most Claude Code workflows, you will not need to change the default.

Switch to Opus 5 Fast (same quality, 2.5x speed, ~2x price) when turnaround time is the constraint and budget allows.

When to Use Fable 5

Fable 5 earns its price premium in a specific scenario: work that requires days-long uninterrupted autonomy with self-correction.

Use Fable 5 for:

• Autonomous agent runs spanning multiple days with minimal human steering
• Problems where choosing the wrong approach is catastrophically expensive
• Workloads where Anthropic explicitly recommends it (per their model documentation)
• Teams that were previously using Fable 5 successfully for long-horizon tasks and want to maintain that baseline before evaluating Opus 5

Fable 5 requires opting into Anthropic’s 30-day data retention policy. If your organization has data handling requirements that preclude this, Opus 5 (which has no mandatory retention) is effectively the only option regardless of performance comparison.

What About Sonnet 5 and Haiku 4.5?

Opus 5’s launch has shifted the model lineup positioning.

Sonnet 5 sits in an increasingly compressed position. For tasks where quality matters, the gap between Sonnet 5 and Opus 5 is often worth the cost. For high-volume, cost-sensitive workloads, Haiku 4.5 remains the right choice.

Practical Model Selection Guide

How to Access Opus 5

API:

import anthropic

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-opus-5",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Your prompt here"}
    ]
)

Claude Code:

Opus 5 is the default model for Claude Max subscribers. To set it explicitly in a session:

/model claude-opus-5

Claude.ai:

Opus 5 is available to Pro and Max subscribers as the top available model in the model selector.

Want Help Choosing the Right Model for Your AI Workflows?

Model selection is one decision in a larger AI strategy. Companies that choose correctly on model, workflow, and architecture compound their advantage. Those that optimize for one variable in isolation often do not.

Phos AI Labs is an embedded AI consulting firm for small and mid-market businesses.

We identify the right AI problems, build the strategy, handle implementation, and train your team until AI is how the business actually runs.

• Strategy before systems: We establish which models, tools, and workflows fit your actual operations before any implementation begins.
• AI Foundations that hold: We install the operating context, decision rules, and configuration standards your team runs on for years.
• Real team training: We build fluency inside your actual workflows, not in staged demos disconnected from your operations.
• Private AI Workspace: We design a company-wide AI environment built around your knowledge base and existing stack.
• AI Implementation: We rebuild the workflows that matter most so AI compounds across your business.
• Honest judgment, every time: We tell you which model is right for your specific workload, not what is newest.
• We stay until it compounds: We are not done when the recommendation is delivered. We are done when the business runs differently.

400+ engagements. Clients include Zapier, Coca-Cola, Medtronic, Dataiku, and American Express.

For Claude-native AI development using the right model for each workload, LOW/CODE Agency is one of the first Anthropic partners worldwide with 10+ CCA-F certified developers on staff.

If you want to get your AI model and workflow decisions right, talk to the team at Phos AI Labs.

Frequently Asked Questions

Is Claude Opus 5 better than Fable 5?

On most published benchmarks, yes. Opus 5 leads on agentic coding (43.3% vs 33.7%), knowledge work (1,861 vs 1,747 Elo), and computer use. Fable 5’s remaining advantage is multi-day long-horizon autonomy.

How much does Claude Opus 5 cost?

$5 per million input tokens and $25 per million output tokens, identical to Opus 4.8. Fable 5 costs $10/$50 per million tokens. Opus 5 Fast costs $10/$50 at 2.5x speed.

What is the model ID for Claude Opus 5?

The model ID is claude-opus-5. Use this in API calls and in Claude Code with /model claude-opus-5.

Does Opus 5 require data retention like Fable 5?

No. Fable 5 requires opting into Anthropic’s 30-day data retention policy. Opus 5 has no mandatory retention requirement, consistent with prior Opus models.

When should I still use Fable 5 over Opus 5?

When your workload requires multi-day autonomous runs with sustained self-correction. Fable 5 remains Anthropic’s pick for the longest-horizon autonomous work. For most other tasks, Opus 5 matches or exceeds Fable 5 at half the price.

What happened to Fable 5 in June 2026?

Fable 5 was taken offline June 12 to July 1, 2026, following a US government export control directive. Anthropic restored global access after approximately two weeks of negotiations.