Claude Opus 5 API: Near-Fable Intelligence at Half the Price (2026 Guide)

On July 24, 2026, Anthropic released Claude Opus 5 — the latest member of the Opus family and the closest Anthropic has come to delivering Fable 5 intelligence at the Opus 4.8 price point. Opus 5 costs exactly the same as its predecessor ($5 per million input tokens, $25 per million output tokens) while more than doubling Opus 4.8's score on Frontier-Bench v0.1 and performing within 0.5% of Fable 5 on CursorBench 3.2 at half the cost per task.

This is the reference guide for developers evaluating Claude Opus 5 for production: the price, the benchmarks where it shines, the benchmarks where it trails Mythos 5 and Fable 5, code examples for the new API features (effort settings, mid-conversation tool changes, automatic fallbacks), and a direct cost comparison against the broader Claude family and competing frontier models. All pricing and benchmark figures in this article are verified from the official Anthropic announcement published July 24, 2026, and the Claude Platform documentation as of late July 2026.

If you are deciding whether to upgrade from Opus 4.8, adopt Opus 5 over Fable 5 for cost savings, or evaluate the entire Claude family's 2026 lineup, this guide has the numbers you need.

Claude Opus 5 Pricing: $5/$25 with Fast Mode Option

Claude Opus 5 is priced at the same rate as Opus 4.8, making it the most cost-effective upgrade in the Opus line's history:

The key takeaway: Opus 5 delivers near-Fable-5 performance at one-third the Fable 5 price. Fast Mode is useful for latency-sensitive applications like chat, code completions, and interactive agents where 2.5x throughput matters more than the 2x price premium.

Benchmarks: Where Opus 5 Dominates

Anthropic published a comprehensive set of benchmark results in the launch announcement. These are the four most impactful for production builders:

Frontier-Bench v0.1: More Than Double Opus 4.8

On Frontier-Bench v0.1 (Anthropic internal coding evaluation), Opus 5 surpasses all other models and more than doubles Opus 4.8's performance at a lower cost per task. This is the broadest measure of software engineering capability in Anthropic's benchmark suite.

CursorBench 3.2: Within 0.5% of Fable 5 at Half the Cost

On CursorBench 3.2 (Cursor's IDE agent benchmark), Opus 5 at max effort performs within 0.5% of Fable 5's peak score — but at approximately half the cost per task. At every effort setting (high, xhigh, max), Opus 5 achieves greater performance at a given cost than all other models.

ARC-AGI 3: Three Times the Next-Best Model

On ARC-AGI 3 (Abstraction and Reasoning Corpus), which tests a model's ability to solve novel problems it has never seen before, Opus 5's score is three times as high as the next-best model. This is the most striking single number in the launch — it suggests Opus 5 has a fundamentally stronger reasoning capability than any other publicly available model.

Zapier AutomationBench: 1.5x Pass Rate of the Next-Best Model

On Zapier AutomationBench (end-to-end business task completion), Opus 5's pass rate is approximately 1.5 times the next-best model for the same cost per task. Even at its lowest effort setting, Opus 5 passes more tasks than any other model at any effort level. Zapier CEO Wade Foster confirmed: Opus 5 topped the leaderboard "without spending more tokens than prior Claude models."

Using Claude Opus 5 with the API

Here is how to call Claude Opus 5 via the Messages API using curl and Python. The model ID is claude-opus-5.

curl Example

curl https://api.anthropic.com/v1/messages   -H "x-api-key: $ANTHROPIC_API_KEY"   -H "anthropic-version: 2023-06-01"   -H "content-type: application/json"   -d '{
    "model": "claude-opus-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Explain how Claude Opus 5 compares to Opus 4.8 in one paragraph."}]
  }'

Python Example with Effort Setting

import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-opus-5",
    max_tokens=2048,
    thinking={
        "type": "enabled",
        "budget_tokens": 4096,
        "effort": "high"
    },
    messages=[
        {"role": "user", "content": "Write a complex Python script to analyze a dataset and generate visualizations."}
    ]
)

print(response.content[0].text)

The effort parameter (new with Opus 5) lets you control how thoroughly the model thinks before responding. Use low or medium for simple tasks to save tokens, high or xhigh for complex reasoning, and max to match the launch benchmarks. At lower effort levels, Opus 5 still outperforms Opus 4.8 at max effort on many tasks — making it both smarter and cheaper to run.

Claude Opus 5 vs Fable 5 vs Opus 4.8: Which Should You Use?

For the vast majority of production use cases — coding assistants, agent frameworks, customer support chatbots, document analysis, data extraction — Opus 5 is the optimal price-performance point. Only use Fable 5 when you need the absolute frontier in cybersecurity, biology research, or tasks where safety restrictions are acceptable. Keep Opus 4.8 configured as a fallback model for requests that trigger Opus 5's safety classifiers (automatic fallbacks are available in the API as a beta feature).

Safety and Alignment: Best-in-Class Opus

Anthropic's pre-deployment automated behavioral audit found Opus 5 to be their most aligned model to date. It adheres to Claude's Constitution better than Opus 4.8, Sonnet 5, or Fable 5 — exhibiting the lowest rates of deceptive behavior and the lowest susceptibility to being tricked into misuse. On the overall misaligned-behavior score, Opus 5 scores 2.3, the lowest of any recent Anthropic model.

Opus 5's cyber classifiers are proportionally less restrictive than Fable 5's — they allow vulnerability identification in source code but block binary-based vulnerability scanning, penetration testing, and exploit generation. For cybersecurity researchers, Anthropic's Cyber Verification Program provides access to a less-restricted version of Opus 5. For biology, Opus 5's safeguards match Opus 4.8's, making it the most capable generally available model for life sciences research.

Verdict: The Obvious Upgrade from Opus 4.8

Claude Opus 5 is the easiest upgrade decision in the Opus family's history: same price, dramatically better performance. There is no scenario where a production builder should choose Opus 4.8 over Opus 5 at equal pricing, unless Opus 4.8 is needed as a safety-classifier fallback (which the API now supports automatically).

The model makes Fable 5 hard to justify for everyday use — Opus 5 matches or approaches Fable 5 on coding, agentic, and knowledge-work benchmarks at one-third the price. Only pick Fable 5 for cybersecurity, advanced biology research, or tasks requiring the absolute frontier.

If you are building with Claude today, the right default model for production is claude-opus-5. It is available now on the Claude API, Claude Code, and Claude.ai.