Claude Opus 5: Price, Benchmarks, and Why It Beats Fable 5 on Cost
Claude Opus 5 is coming next week with near-Fable 5 performance for a fraction of the cost. Support for Fable 5 ends July 19.
Updated October 21, 2026. Claude Opus 5 is no longer coming next week. Anthropic released it on July 24, 2026, at the same price as Opus 4.8 and roughly half the price of Fable 5. It is now the default model on Claude Max and the strongest model on Claude Pro.
The short version: Opus 5 closes most of the gap to Fable 5 on coding and knowledge work while costing far less per task. It is not the absolute frontier on every benchmark, but it is the model most teams can actually afford to run all day.
Claude Opus 5 price
| Item | Price |
|---|---|
| Input | $5 per million tokens |
| Output | $25 per million tokens |
| 5-minute cache write | $6.25 per million tokens |
| 1-hour cache write | $10 per million tokens |
| Cache read | $0.50 per million tokens |
| Batch API | 50% off input and output |
| Fast mode | 2x base price, about 2.5x speed |
The headline is not a price cut. Opus 5 costs the same as Opus 4.8. The saving comes from needing fewer retries and fewer tokens to finish a task, which is why Anthropic compares cost per task rather than cost per token.
Benchmarks: where Opus 5 lands
Anthropic says Opus 5 sets a new state of the art on Frontier-Bench and GDPval-AA, two evaluations that focus on real coding and knowledge work rather than trivia. Independent numbers from Decrypt and OpenRouter broadly match that claim.
- Frontier-Bench v0.1: 43.3% for Opus 5, 33.7% for Fable 5, 34.4% for GPT-5.6 Sol.
- ARC-AGI-3: 30.2% for Opus 5, 7.8% for GPT-5.6 Sol. Fable 5 was not tested.
- GDPval-AA v2: 1,861 Elo for Opus 5, 1,747 for Fable 5, 1,736 for GPT-5.6 Sol.
- SWE-bench: 97.0% on the Vals AI run.
- GPQA Diamond: 93.7% at xhigh effort on Artificial Analysis.
Treat vendor benchmarks with care. Anthropic ran Frontier-Bench internally, and effort settings change the scores. The direction is still clear: Opus 5 is competitive with Fable 5 on coding and agentic work at a much lower price.
What changed from Opus 4.8
Opus 5 is better at long software tasks, tool use, and scientific reasoning. Anthropic highlights a 10.2 percentage point gain on an internal organic chemistry benchmark and a 7.7 point gain on protein variant prediction. On Zapier’s AutomationBench, it completed an account-health workflow end to end that earlier models failed.
The model also has an effort setting. You can run it at low effort for cheaper, faster responses or at max effort for harder problems. That control matters more than a single benchmark score because most production workloads do not need maximum reasoning on every request.
Opus 5 vs Fable 5
Fable 5 remains the more expensive frontier model, and it is still ahead on some safety and cybersecurity evaluations. Opus 5 is behind Mythos 5 on cybersecurity tasks, according to Anthropic. For general coding, research, and business automation, the gap is small enough that price becomes the deciding factor.
If your workload is cost-sensitive, start with Opus 5. If you need the absolute best result on a narrow, high-value task, test Fable 5 and compare cost per completed task, not cost per token.
Where to use it
Opus 5 is available through the Claude API, Amazon Bedrock, and Google Cloud’s Vertex AI. It is also available in Claude apps. On subscriptions, it is the default on Claude Max and the strongest model on Claude Pro.
Good first workloads include coding agents, multi-step research, document review, and workflow automation. The model is designed to be proactive, which helps when a task has many small steps but can be risky when you want strict, narrow behavior. Give it clear boundaries and test it before letting it act without review.
Frequently asked questions
When did Claude Opus 5 launch?
July 24, 2026. It replaced the earlier expectation of an August release.
How much does Claude Opus 5 cost?
$5 per million input tokens and $25 per million output tokens. Fast mode costs twice as much. Batch requests get a 50% discount.
Is Claude Opus 5 free?
No. It is available on paid Claude plans and through paid API access. Claude Pro and Max subscribers get access under their plan limits.
Is Opus 5 better than Fable 5?
It is better on several coding and knowledge-work benchmarks and much cheaper. Fable 5 still wins on some frontier and cybersecurity evaluations. The right choice depends on the task and the budget.
Does Opus 5 replace Opus 4.8?
Yes for most users. It costs the same and performs better. Opus 4.8 remains available as a legacy model for teams that need a fixed version.
The bottom line
Claude Opus 5 is the model Anthropic should have led with. It gets close to Fable 5 on the work most people actually do, costs half as much per task, and is available everywhere Claude is sold. The launch ended months of speculation about Fable 5 access. For teams building coding agents or document workflows, Opus 5 is now the default to beat.
