Claude Opus 5 is Anthropic's new flagship model, released July 24, 2026, and it's already live in Claude.ai, Claude Code, Claude Cowork, and the API under the model ID claude-opus-5. It's priced the same as its predecessor — $5 per million input tokens and $25 per million output tokens — while landing benchmark scores that, on a couple of tests, beat Anthropic's more expensive model.
Short answer: Claude Opus 5 launched July 24, 2026, priced at $5/M input and $25/M output tokens — the same as Opus 4.8. It has a 1M-token context window, a new effort toggle (low to max) for balancing speed and quality, and it's the default model on Claude Max and the strongest option on Claude Pro. Access it at claude.ai, in Claude Code, or via the API.
In my testing, I pulled the pricing and benchmark figures below straight from Anthropic’s own launch post and cross-checked the context window and effort-level details against the current models overview docs, because model launches tend to get restated inaccurately by aggregator sites within a day or two. Here's what Opus 5 actually costs, how it stacks up against Anthropic's other current models, and the exact steps to start using it — whether you're chatting in the app or calling the API.
What you'll need
For casual use, you just need a free claude.ai account — no card required, and Opus 5 is available to try from the model picker the moment you sign in. Anthropic hasn't staged this rollout by tier the way some launches go; it's live across the consumer app, Claude Code, and Claude Cowork already. If you want it as your default model rather than picking it per chat, that currently means being on Claude Max, since Anthropic set Opus 5 as Max's default and merely the "strongest available" option on Pro. For the API path, you need an account at console.anthropic.com, a payment method on file, and a rough token budget in mind — at $5/M input tokens, a 10,000-word document (roughly 13,000 tokens) costs about 6.5 cents to send once, before any output.
Step-by-step: Claude Opus 5
1. Check which surface you're using it from
Claude.ai for chat, Claude Code for terminal-based coding work, Claude Cowork for the newer collaborative workspace product, or the API for anything you're building yourself. Opus 5 is available on all four as of launch, so the choice comes down to what you're doing, not whether the model is accessible yet.
2. Confirm the model picker actually shows Opus 5
Open the model dropdown in claude.ai or Claude Code and look for "Opus 5" specifically, not just "Opus." Anthropic keeps older Opus versions (4.8, 4.7, 4.6) available on paid plans for a while after a new release, so it's easy to be running an older model without noticing if you don't check.
3. Pick your plan based on how often you'll use it
Free gives you access to try Opus 5, but usage caps hit fast at this model's price point. Pro ($17–$20/month) makes Opus 5 the strongest model available to you, though not the default. Max (from $100/month) sets Opus 5 as the default for every new chat and gives you 5x–20x the usage ceiling of Pro — the plan to pick if you'll lean on it daily.
4. Get an API key if you're building anything
Sign in at console.anthropic.com, generate a key, and call the model as claude-opus-5 in your requests. Its reliable knowledge cutoff is May 2026, so anything more recent needs a web search tool attached to the call — same rule as every other current Claude model.
5. Set the effort level instead of leaving it on default
Opus 5 ships with an effort parameter — low, medium, high, xhigh, or max — that trades speed and cost against how much the model reasons before answering. It defaults to high on the API and in Claude Code. In my testing, dropping to low on straightforward tasks (rewriting a paragraph, a simple regex) barely changed the output but noticeably cut wait time, while pushing to xhigh on a multi-file refactor produced a visibly more careful diff at the cost of a slower response.
6. Turn on fast mode if latency matters more than cost
On the API, fast mode runs Opus 5 at roughly 2.5x its default speed, priced at double the base rate. It's worth it for interactive tools where a user is staring at a spinner; it's not worth it for batch jobs where nobody's watching the clock.
Example prompts you can copy
These lean on what Opus 5 is specifically built for — agentic coding and enterprise knowledge work — rather than generic chat requests:
- "Here's my current codebase structure [paste or attach]. Refactor the auth module to use the new session pattern, and don't touch anything outside that module." (Tests agentic scoping — does it stay in its lane.)
- "Review this pull request for bugs, not style. Flag anything that could break in production, and rate your confidence on each one." (Good use of the effort toggle — try it at high vs. xhigh and compare.)
- "I'm building [type of workflow] that needs to call [API/tool]. Write the integration, then write a test that would catch the most likely failure mode." (Coding plus self-critique in one pass.)
- "Summarize this 60-page report in one page for an executive audience, then list the three numbers in it I should verify before repeating them externally." (Tests the 1M-token context window on a real document.)
- "Compare these two vendor proposals line by line and tell me what's missing from each, not just what's there." (Knowledge-work task Anthropic specifically built Opus 5 for.)
Common mistakes to avoid
The first mistake I'd flag: assuming "Opus" in the model picker means Opus 5. Anthropic left 4.8, 4.7, and 4.6 available on paid plans, and they show up in the same dropdown — check the version number, not just the family name. Second, running everything at the default high effort setting when a task doesn't need it; low or medium handles routine writing and simple code edits fine, and you're paying the same per-token rate regardless of effort level, so higher effort mostly costs you time, not just money. Third, expecting Opus 5 to know about anything after its May 2026 training cutoff without a search tool attached — it will reason confidently from stale information instead of flagging the gap. Fourth, defaulting to Opus 5 for tasks Sonnet 5 handles just as well at a third of the input cost; Opus 5 is genuinely strongest on complex agentic coding and enterprise workflows, not every single request. Fifth, turning on fast mode for background or batch work — you're paying double the base rate for speed nobody's waiting on.
Claude Opus 5 vs. Anthropic's other current models
| Model | Input $/M tokens | Output $/M tokens | Context window | Notable |
|---|---|---|---|---|
| Claude Opus 5 | $5.00 | $25.00 | 1M tokens | Default on Claude Max; within 0.5% of Fable 5 on CursorBench 3.2 at half the cost |
| Claude Opus 4.8 | $5.00 | $25.00 | 1M tokens | Previous Opus flagship, still available on paid plans |
| Claude Fable 5 | $10.00 | $50.00 | 1M tokens | Anthropic's most capable model overall; Opus 5 beats its OSWorld 2.0 score at roughly a third of the cost |
| Claude Sonnet 5 | $2.00 (intro, through Aug 31, 2026) | $10.00 (intro) | 1M tokens | Cheapest current-gen model; rises to $3/$15 per M after the intro period |
Pricing confirmed against Anthropic’s own launch post and current models docs on 2026-07-27. Fable 5 stays the higher-capability option for the hardest long-horizon tasks, but on Anthropic's own benchmark set, Opus 5 closes a lot of that gap at half to a third of the per-task cost.
Tools that make this easier
If you're deciding whether to switch your default model or just try Opus 5 alongside what you already use, my how to use Claude AI guide covers the account and plan setup this article assumes you already have. If you're weighing Claude against the other major assistants, how to use ChatGPT and how to use Gemini cover the equivalent setup on each. For coding specifically — Opus 5's strongest category — see how it stacks up in best AI tool for code, or check the wider field in ChatGPT alternatives for coding if none of the big three fit your workflow. I ran the same kind of pricing-and-benchmark check on xAI's current flagship in Grok 4.5: pricing, benchmarks, and setup guide, useful if you're comparing across vendors rather than just within Anthropic's lineup. My starter kit for AI is the place to start if you haven't set up any model accounts yet, and how we test AI tools explains how I verify numbers like these before publishing them.
Where this leaves you
Claude Opus 5 is a real upgrade, not a repricing exercise — same $5/$25 rate as Opus 4.8, but Anthropic's own benchmark set shows it closing most of the gap to the pricier Fable 5 on coding and computer-use tasks, sometimes at a third of the cost. The effort toggle is the feature I'd actually use day to day: dropping to low on routine work saves real wait time, and bumping to xhigh on something that matters is a one-parameter change, not a different model to learn. If you're on Claude Max, you already have it as your default. If you're on Pro or Free, it's one click away in the model picker — worth trying on your next coding or long-document task before assuming your current setup is good enough.
Last updated: July 27, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Frequently Asked Questions
Claude Opus 5: is it free?
You can access Opus 5 on Claude's free plan, but usage caps hit quickly at this model's price point since Free's overall usage allowance is the smallest tier. The API isn't free either way — it's pay-per-token at $5/M input and $25/M output, billed to a card on file at console.anthropic.com.
How long does it take to start using Claude Opus 5?
Through claude.ai or Claude Code, a couple of minutes — open the model picker and select it if you're not already on Max, where it's the default. Through the API, expect 10–15 minutes to create an account, generate a key, and send a first test call with claude-opus-5 as the model string.
What is the easiest way to try Claude Opus 5?
Open claude.ai, select Opus 5 from the model picker, and run a real task through it — a coding change or a long document — rather than a generic chat prompt. That's the fastest way to judge whether it's worth switching your default model.
Is Claude Opus 5 better than Claude Fable 5?
Not across the board — Fable 5 is still Anthropic's most capable model and costs twice as much per token. But on Anthropic's own benchmarks, Opus 5 gets within 0.5% of Fable 5's score on CursorBench 3.2 at half the cost, and it beats Fable 5's OSWorld 2.0 result at roughly a third of the cost, so for a lot of agentic coding work, Opus 5 is the better value even if it's not the higher ceiling.
How does Claude Opus 5 compare to Claude Sonnet 5 on price?
Sonnet 5 is meaningfully cheaper — $2/M input and $10/M output during its introductory pricing through August 31, 2026, versus Opus 5's $5/M and $25/M. Sonnet 5 is the better default for routine work; Opus 5 earns its higher price on complex agentic coding and enterprise knowledge tasks specifically, per Anthropic's own positioning.