Code with Claude Opus 4.7
in AdaL CLI
Anthropic's most capable coding model. A meaningful jump over Opus 4.6 with 3× sharper vision, xhigh effort control, and the reliability to handle your hardest engineering work autonomously.
No credit card required · $5/M input · $25/M output tokens
Why developers choose Opus 4.7
The most powerful upgrade for complex, long-running professional tasks
Advanced Software Engineering
Handles complex, long-running tasks with rigor and consistency. Devises ways to verify its own outputs before reporting back.
3× Higher Resolution Vision
Accepts images up to 2,576px on the long edge (~3.75MP) — 3× more pixels than prior models. Reads dense screenshots and complex diagrams.
Precise Instruction Following
Takes instructions literally and follows them rigorously. A significant leap in reliability for prompt-driven workflows.
New xhigh Effort Level
Fine-grained control between high and max effort. Deeper reasoning on hard problems with predictable latency tradeoffs.
Built-in Cyber Safeguards
First model with real-time safeguards that detect and block prohibited cybersecurity uses, while supporting legitimate security research.
Benchmark Leader
70% CursorBench (+12pp over Opus 4.6), 98.5% XBOW visual acuity, state-of-the-art on finance agent evals and GDPval-AA.
Opus 4.7 Benchmark Performance
State-of-the-art across coding, vision, and agentic benchmarks
| Benchmark | Opus 4.7 | Opus 4.6 | Sonnet 4.6 | Gemini 3.1 Pro | GPT-5.4 |
|---|---|---|---|---|---|
| Agentic coding (SWE-bench Verified) | — | 80.8% | 77.2% | 76.2% | — |
| Agentic terminal coding (Terminal-Bench 2.0) | — | 65.4% | 51.0% | 56.2% | — |
| IDE coding (CursorBench) | 70.0% | 58.0% | — | — | — |
| Agentic search (BrowseComp) | — | 84.0% | 43.9% | 59.2% | — |
| Finance agent (Scale AI) | SOTA | — | — | — | — |
| Visual acuity (XBOW) | 98.5% | 54.5% | — | — | — |
Scores from Anthropic's announcement and partner testimonials. "—" = not reported. SOTA = state-of-the-art.
What teams are saying
Early-access feedback from industry leaders
"On CursorBench, Opus 4.7 is a meaningful jump in capabilities, clearing 70% versus Opus 4.6 at 58%."
Cursor
"Opus 4.7 works coherently for hours, pushes through hard problems rather than giving up, and unlocks deep investigation work we couldn't run before."
Cognition (Devin)
"It's phenomenal on one-shot coding tasks, more correct and complete than Opus 4.6, and noticeably more honest about its own limits."
Vercel
"+14% over Opus 4.6 at fewer tokens and a third of the tool errors. First model to pass our implicit-need tests."
Notion
"Same quality at lower cost — more efficient and precise. It really feels like a better coworker."
Replit
"Up to 10% better without the regressions we've come to expect from very agentic models."
Bolt
How to use Claude Opus 4.7
Get started in minutes
Follow the Quickstart
Head to our docs to get set up in minutes.
Sign up (free)
Create your account — takes less than a minute.
Select Claude Opus 4.7
Use /model to pick Claude Opus 4.7 from the model selector.
Ready to code with the most powerful AI model?
Experience state-of-the-art coding performance with Claude Opus 4.7 in AdaL CLI.
Get Started with AdaL