Claude Opus 5 is Anthropic's current Opus-tier model: a 1M-token context window, 128K maximum output, adaptive reasoning on by default, and pricing of $5 per million input tokens and $25 per million output — the same as Opus 4.8 before it, and half what Claude Fable 5 charges.
This is the hub for everything we have written about it: what it costs, what it is good at, how it compares with its siblings and its competitors, and the prompt habits that change when you move to it. If you are new to Claude generally, start with our Claude review instead.
Specifications
| Spec | Claude Opus 5 |
|---|---|
| Model ID (API) | claude-opus-5 |
| Context window | 1M tokens (default and maximum) |
| Max output | 128K tokens per response |
| Price | $5 input / $25 output per million tokens |
| Fast mode | $10 / $50, up to 2.5x more output tokens per second (Claude API only) |
| Reasoning | Adaptive thinking on by default |
| Effort levels | low, medium, high, xhigh, max (default high) |
| Vision | High resolution, 2,576 px on the long edge |
| Where to run it | Claude API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry |
What Claude Opus 5 is actually for
Opus 5 is the successor to Opus 4.8 in the Opus line, positioned for complex agentic coding, deep reasoning, and long-horizon work — the tasks where a model has to stay coherent for a long time and finish what it started. It is deliberately not the right default for everything: routine, high-volume work belongs on Claude Sonnet 5 or Haiku 4.5, and the very hardest multi-day autonomous runs may justify Claude Fable 5 at double the price.
Three characteristics define working with it day to day:
- It finishes. Multi-file features and large refactors come back complete rather than stubbed.
- It verifies itself. Which means the "double-check your work" instruction in your prompt library is now actively harmful.
- It writes long. Verbosity is the one thing you will almost certainly need to prompt against.
Use cases
Comparisons
| Comparison | The question it answers |
|---|---|
| Claude Opus 5 vs Claude Opus 4.8: What Actually Changed | Claude Opus 5 against Claude Opus 4.8 |
| Claude Opus 5 vs Claude Sonnet 5: When Is Opus Worth 1.7x the Price? | Claude Opus 5 against Claude Sonnet 5 |
| Claude Opus 5 vs Claude Haiku 4.5: Frontier Reasoning vs Volume Economics | Claude Opus 5 against Claude Haiku 4.5 |
| Claude Opus 5 vs Claude Fable 5: Which Frontier Claude Do You Need? | Claude Opus 5 against Claude Fable 5 |
| Claude Opus 5 vs Claude Opus 4.7: Should You Skip a Generation? | Claude Opus 5 against Claude Opus 4.7 |
| Claude Opus 5 vs GPT-5.6 Sol: The Real Price Comparison | Claude Opus 5 against GPT-5.6 Sol |
| Claude Opus 5 vs ChatGPT: Which Should You Actually Subscribe To? | Claude Opus 5 against ChatGPT |
| Claude Opus 5 vs Gemini: Reasoning Depth vs Google Integration | Claude Opus 5 against Gemini |
Specs, pricing and setup
| Guide | What it covers |
|---|---|
| Claude Opus 5 Pricing Explained (2026) | Claude Opus 5 costs $5 per million input tokens and $25 per million output. What that means in practice, plus every way to pay less. |
| Claude Opus 5 Context Window: 1M Tokens, Explained | What a 1M-token context window means in pages, code, and cost — and how to use it without wasting money. |
| How to Access Claude Opus 5: Every Route in 2026 | Apps, API, Claude Code, and the three cloud platforms — every way to use Claude Opus 5, and which one fits you. |
| Claude Opus 5 API: A Practical Getting-Started Guide | Model ID, parameters, and the four API defaults that surprise people migrating to Claude Opus 5. |
| Claude Opus 5 Effort Levels: Which One Should You Use? | low, medium, high, xhigh, max — what the Claude Opus 5 effort parameter actually changes, and how to pick. |
| Claude Opus 5 Extended Thinking: On by Default, and Why That Matters | Adaptive thinking on Claude Opus 5 — what changed from Opus 4.8, when to disable it, and the two failure modes if you do. |
| Claude Opus 5 Fast Mode: Double the Price, Up to 2.5x the Speed | What fast mode does on Claude Opus 5, what it costs, where it's available, and when it's actually worth it. |
| Claude Opus 5 Rate Limits: What to Check Before You Scale | Claude Opus 5 uses its own rate-limit bucket and is excluded from Priority Tier. What that means for production planning. |
| Claude Opus 5 on AWS, Google Cloud and Azure | Running Claude Opus 5 on Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry — model IDs and feature differences. |
| How to Cut Claude Opus 5 Costs Without Downgrading | Prompt caching, batching, effort tuning, and routing — four levers that cut Claude Opus 5 spend while keeping the model. |
| Upgrading to Claude Opus 5: A Migration Checklist | Moving from Opus 4.8 or earlier to Claude Opus 5 — what breaks, what to delete from your prompts, and what to re-tune. |
| Claude Opus 5 Limitations: What It Doesn't Do Well | An honest list of Claude Opus 5's weak points — verbosity, scope creep, cost, refusals, and the things it simply cannot do. |
Where Claude Opus 5 sits in the line
| Model | Price (per 1M tokens) | Use it for |
|---|---|---|
| Claude Fable 5 | $10 / $50 | The hardest long-horizon autonomous work |
| Claude Opus 5 | $5 / $25 | Complex coding, deep reasoning, agents |
| Claude Opus 4.8 | $5 / $25 | The previous Opus; Priority Tier workloads |
| Claude Sonnet 5 | $3 / $15 | The everyday default for most production traffic |
| Claude Haiku 4.5 | $1 / $5 | High-volume, well-specified, latency-sensitive tasks |
Most teams that get this right do not pick one model. They route: a cheap model by default, Opus 5 on escalation, and a measured evaluation set deciding where the boundary sits.
How to get it
Four routes — the Claude apps, Claude Code, the Claude API, and the three major cloud platforms. Our access guide covers which one fits you and what each gives up.
Ready to go deeper?
Compare AI chatbotsFrequently Asked Questions
What is Claude Opus 5?
Claude Opus 5 is Anthropic's current Opus-tier model and the successor to Claude Opus 4.8. It has a 1M-token context window, 128K maximum output, adaptive reasoning on by default, and costs $5 per million input tokens and $25 per million output. It is positioned for complex agentic coding, deep reasoning, and long-horizon work.
How much does Claude Opus 5 cost?
$5 per million input tokens and $25 per million output on the Claude API — the same as Opus 4.8, and half of Claude Fable 5's $10/$50. Fast mode doubles both rates, the Batch API halves them, and cached input reads cost roughly a tenth of the standard rate.
Is Claude Opus 5 better than Claude Fable 5?
No — Fable 5 sits above the Opus tier and is Anthropic's most capable widely released model, at twice the price. Opus 5 is the better default for most work; Fable 5 is the escalation choice for the hardest multi-day autonomous runs. Fable 5 also requires 30-day data retention, which rules it out for zero-retention organisations.
What is Claude Opus 5 best at?
Agentic coding on difficult tasks — multi-file features, large refactors, unfamiliar codebases — plus code review, long-document analysis using the full 1M-token window, multi-sheet spreadsheet and slide generation, and long autonomous agent runs. The gap over cheaper models is smallest on simple, single-turn tasks.
