Claude Opus 5 is out: what changed and which model to pick (July 2026)
On July 24 Anthropic released Claude Opus 5: it is now the main model, and Opus 4.8 moved to Legacy. Here is what changed and which model to pick per task.
On July 24, 2026 Anthropic released Claude Opus 5. It is now the main working model of the lineup: the previous Opus 4.8 moved to the Legacy section of the official model table. If you picked your model based on our earlier breakdowns, the picture has changed, so let's update it.
What exactly happened?
The official Anthropic model table now lists four current models: Fable 5, Opus 5, Sonnet 5 and Haiku 4.5. Opus 4.8, together with 4.7 and 4.6, sits in the collapsed Legacy models block: they still work, but Anthropic is clearly betting on the new ones.
Opus 5 is positioned as the model for complex work: long multi-step reasoning, large documents, code. In the API it is priced at $5 per million input tokens, half the price of Fable 5 at $10. The most capable "everyday" tier just got more affordable.
The release is confirmed by the official announcement at anthropic.com/news/claude-opus-5 and independently by Axios.
Which model should you pick per task?
The logic hasn't changed. The names on the shelves have.
| Task | Model |
|---|---|
| Quick small things: rewrite an email, short answer, a draft | Haiku 4.5 |
| Everyday work: texts, tables, breakdowns, most tasks | Sonnet 5 |
| Complex and multi-step: strategy, large documents, code, automations | Opus 5 |
| Extremely hard cases where the others fell short | Fable 5 (on the Pro plan now via credits, details here) |
If you don't want to think about the choice at all, keep whatever the selector shows by default and switch manually only when an answer disappoints. The effort rule stays the same: higher effort means deeper thinking and slower answers.
What to do right now?
- Open the model selector in Claude (at the bottom of the input box) and see what's new there. The updated list arrives on its own, nothing to install.
- Take your hardest recurring task and run it through Opus 5. Compare with what the previous model gave you.
- If you work through the API, check which model your automations are pinned to: model names change, and legacy ones get retired over time.
What else did Anthropic ship in July?
Two updates that are easy to miss.
Reflect. A panel in Claude's settings that shows what you've been doing with Claude: topics, active hours, usage over time. Useful for seeing where your AI time actually goes.
Live artifacts. A published artifact (a page Claude builds for you) can now pull fresh data from your connected services every time it opens, instead of showing a snapshot from the moment it was made. Public links arrived too. That turns artifacts into proper mini dashboards: build a report, send the link to a partner, and the numbers are always today's. Source: Claude weekly digest, week 29.
We show how to try all of this hands-on in our free guides and the first lesson with no signup.
AGINE Academy is an independent product, not affiliated with Anthropic. Claude is a product of Anthropic.
Questions
Opus 5 replaced Opus 4.8 as the main model for complex tasks: the official Anthropic model table lists it as current, while 4.8 moved to Legacy. In the API Opus 5 is priced at $5 per million input tokens. The old model still works, but development goes into the new one.
Yes. Opus 5 is available on paid Claude plans on the web, in the app and via the API. Open the model selector at the bottom of the input box and pick it. Nothing extra to install or pay beyond the subscription.
Nothing urgent: it works, but it is listed as legacy. Use Opus 5 for new tasks, and if you have API automations, migrate them to current models in advance: Anthropic retires legacy models over time.