How do I use Claude Opus 5 in Cursor?

Update Cursor to the latest version, open Settings → Models, and select Claude Opus 5 (it became the default Opus after its July 24, 2026 launch). If it's not listed, add your Anthropic API key under Settings → Models → API Keys. Opus 5 is priced the same as Opus 4.8 ($5/$25), so switching doesn't change your rate card.

How do I switch Claude Code to Opus 5?

Run /model in Claude Code and pick Claude Opus 5, or set the model in your config/environment. Opus 5 is the recommended default for agentic coding. You can also set an effort level (low/medium/high/xhigh/max) to trade quality against output-token cost per call.

What's the model ID for Claude Opus 5?

Use the Opus 5 model identifier from Anthropic's docs (e.g., the claude-opus-5 family ID) in your API calls, or select Claude Opus 5 in Cursor/Claude Code. It supports a 1M-token context and 128K max output, with a lower 512-token prompt-cache minimum.

Quick Answer

How to Switch to Claude Opus 5 in Cursor & Code

Published: July 27, 2026

The Short Answer

Claude Opus 5 (July 24, 2026) is the new default Opus at the same $5/$25 price as Opus 4.8, so switching is low-risk. Update your client, select Opus 5, set an effort level, and re-baseline prompt caching.

In Cursor

Update Cursor to the latest build.
Open Settings → Models.
Select Claude Opus 5 (default Opus post-launch). If missing, add your Anthropic API key under Settings → Models → API Keys.
(Optional) Set a lower effort/thinking level for cheaper routine edits.

In Claude Code

Run /model and choose Claude Opus 5, or set it in your config/environment.
Set an effort level — start at high; drop to medium for routine work.
Opus 5’s stronger self-correction changes multi-turn behavior — re-run a couple of familiar tasks to recalibrate.

Via the API

Swap your model ID to the claude-opus-5 family (see Anthropic docs for the exact string).
Keep the 1M context / 128K output envelope — same as 4.8.
Take advantage of the 512-token prompt-cache minimum (down from 1,024) so shorter prompts cache.
Optionally adopt mid-conversation tool changes (beta) to add/remove tools between turns without resending the full tool list.

Effort Levels (new in Opus 5)

Effort	Use for
low / medium	routine edits, cheaper calls
high	default for real coding
xhigh / max	hardest reasoning, multi-repo refactors

Higher effort → more output tokens → higher cost. Tune per workload.

Safe Migration from Opus 4.8

Change the model ID; leave everything else fixed.
Regression-test agent/multi-turn flows.
Keep Opus 4.8 or Sonnet 5 pinned as a fallback while you validate.
Re-baseline any output-token cost dashboards — effort level now drives token count.

Sources

Claude docs — what’s new in Opus 5: platform.claude.com/docs/en/about-claude/models/whats-new-opus-5
Anthropic — Claude Opus 5 (July 24, 2026): anthropic.com/news/claude-opus-5
OpenRouter — Opus 5: openrouter.ai/anthropic/claude-opus-5