Claude Opus 5 Pricing: Near-Fable Intelligence at Half the Cost for Heavy AI Users

Anthropic released Claude Opus 5 on July 24, 2026, and the numbers tell a clear story for heavy AI users: near-frontier intelligence at half the cost of Fable 5.

At $5 per million input tokens and $25 per million output tokens — the same price as Opus 4.8 — Opus 5 delivers performance that on several benchmarks matches or exceeds Fable 5 (priced at $10/$50 per MTok). For teams spending $5,000 to $50,000 per month on API usage, the math is compelling.

What Opus 5 brings to the table

Opus 5 is not a minor refresh. Anthropic’s own benchmarks show it surpassing Fable 5 on Frontier-Bench v0.1, GDPval-AA, and ARC-AGI 3. On CursorBench 3.2, it scores within 0.5% of Fable 5 at max effort but at half the cost per task. On OSWorld 2.0, it beats Fable 5’s best result at just over a third of the cost.

For coding workflows, the model excels at self-verification. Given a drawing of a machine part with no direct way to view it, Opus 5 wrote its own computer vision pipeline to extract geometry and reconstructed the full 3D model. A competing model failed this task across five attempts.

The cost advantage for heavy API users

The pricing comparison breaks down like this:

• Claude Opus 5: $5 input / $25 output per MTok (same as Opus 4.8, roughly half of Fable 5)
• Claude Fable 5: $10 input / $50 output per MTok (the frontier model, gated behind stricter access controls)
• Claude Sonnet 5: $3 input / $15 output per MTok (intro pricing at $2/$10 ends August 31)
• GPT-5.6 Sol: $5 input / $30 output per MTok (OpenAI’s top-tier model)
• GPT-5.6 Terra: $2.50 input / $15 output per MTok

For a team running 100 million input tokens and 20 million output tokens per month on coding and knowledge work, switching from Fable 5 to Opus 5 cuts the monthly API bill from $1,500 to $750. That is a 50% reduction with benchmark performance that on some tasks actually exceeds Fable 5.

Fewer restrictions, lower compliance costs

Opus 5 comes with notably lighter restrictions than Fable 5. It is not subject to the 30-day data retention policy that covers Fable and Mythos, removing a compliance concern for many enterprises. The safety classifiers are expected to intervene around 85% less often than they do for Fable 5, reducing the friction of flagged requests.

When classifiers do flag a request, it falls back to Opus 4.8 by default rather than being blocked outright. This fallback mechanism is available on the API too, meaning production pipelines stay running even when safety filters trigger.

Effort-based pricing gives you control

Opus 5 introduces meaningful effort tiers. At its lowest effort setting, it uses fewer tokens while still passing more tasks on Zapier AutomationBench than any competing model at any effort level. At max effort, it approaches Fable 5 territory at half the cost.

This matters for heavy users who can route simpler tasks to low-effort Opus 5 and reserve max effort for the hardest problems. The model’s architecture makes this tradeoff more efficient than previous generations.

Fast mode for speed-sensitive workflows

Opus 5 is also available in Fast mode, running around 2.5 times the default speed at twice the base price. At $10/$50 per MTok, Fast mode costs the same as standard Fable 5 but delivers Fable-competitive results at higher throughput. For latency-sensitive applications like real-time code review or customer-facing agents, this is a practical alternative.

How Opus 5 fits the Claude lineup

Opus 5 sits in a sweet spot. It is the default model on Claude Max (replacing Opus 4.8) and the strongest model on Claude Pro. Fable 5 remains the absolute frontier but comes with access gates and higher pricing. Sonnet 5 handles everyday workloads at a lower cost tier with introductory pricing ending August 31.

For heavy AI users, the practical decision tree is straightforward: if your task needs the absolute maximum capability and you have budget for it, use Fable 5. For everything else — coding, analysis, research, knowledge work — Opus 5 delivers comparable results at half the token cost.

What this means for your monthly budget

Consider a typical heavy user running Claude Code with frequent subagent calls. At 50 million input tokens and 10 million output tokens per day, the daily cost difference is:

• Fable 5: $500 input + $500 output = $1,000 per day
• Opus 5: $250 input + $250 output = $500 per day
• Opus 5 (Fast mode): $500 input + $500 output = $1,000 per day (same as Fable 5 but 2.5x faster)

For teams on the fence about moving workloads to Fable 5, Opus 5 removes the cost objection. The performance gap is narrow enough on most real-world tasks that the savings justify the switch.

Bottom line

Claude Opus 5 is the most cost-effective model in Anthropic’s lineup for heavy AI users who need frontier-adjacent intelligence without the frontier price tag. At $5/$25 per MTok with no data retention policy, automatic fallbacks, and effort-based optimization, it is a practical default for production workloads.

The launch also signals Anthropic’s pricing strategy: keeping Opus as the workhorse tier while positioning Fable as the premium option. For heavy users managing $10,000+ monthly API bills, Opus 5 is the model to build around for the second half of 2026.