Anthropic released a new flagship model on July 24: Claude Opus 5. No free-for-everyone rollout like Sonnet 5 got three weeks earlier. This time the message is different: if you want the best, you pay for it.
What's new in Opus 5?
Opus 5 is currently Anthropic's strongest officially available model, and early reports put it close to the capabilities of Fable 5 — the experimental "myth" model that briefly appeared in June before the US government ordered it shut down again. The difference: Opus 5 is staying. And it costs roughly half of what a model in this class usually demands.
The standout new feature is an effort dial. You — or developers via the API — can set the model to think with low, medium, or high effort. More effort means more thorough answers, but also higher cost and longer wait times. By default, Opus 5 runs on "high" — which honestly says a lot about how confident Anthropic is in this model.
Quick reality check: what do the model names even mean?
If Claude's naming scheme confuses you, you're not alone. Roughly: Haiku is the fast, cheap tier. Sonnet is the balanced default — currently free on claude.ai. Opus is the flagship, built for the hardest tasks. Fable sits a notch above all of them as an experimental showcase model that only surfaces occasionally. Opus 5 slots in as the reliable, always-on top tier.
What's free — and what actually costs money?
Short version: nothing here is free. Opus 5 is strictly for paying users.
- Free Claude access: no Opus 5, Sonnet models only.
- Claude Pro (currently around $20/month): Opus 5 is now the strongest model included.
- Claude Max (currently around $100/month): Opus 5 is the new default.
- Via the API: currently in the range of $5 per million input tokens and $25 per million output tokens — relevant for developers, fairly abstract for casual chat users.
For context: a million tokens is roughly 750,000 words. A single conversation on claude.ai will rarely cost more than a few cents — but anyone building Opus 5 into an app with heavy traffic will feel the bill fast.
What can I try right now?
Already on Claude Pro or Max? Just switch the model picker on claude.ai — Opus 5 is already there. No setup, no waitlist.
No paid plan? Claude Sonnet 5 remains your best free option for now, and Anthropic itself calls it "close to top-tier." The gap to Opus 5 exists, but for everyday tasks — emails, research summaries, explaining code snippets — it's barely noticeable.
Is this for me — or just for developers?
Honestly? For beginners, Opus 5 changes almost nothing. The target audience is companies and developers automating complex tasks who need to weigh cost against quality precisely — that's exactly what the effort dial is for. If you use AI for everyday things, free Sonnet 5 is still the right answer.
What's interesting anyway: Anthropic is positioning itself against the industry's "bigger and pricier" trend. A model that comes close to the top while costing half as much is a statement, not a footnote. If you work with AI APIs professionally, keep Opus 5 on your radar. Everyone else can stay comfortably on free Sonnet — and maybe remember the name for the next AI trivia round.