Claude Opus 4.7
Claude Opus 4.7 is Anthropic’s most capable generally available model. It is much better than Opus 4.6 at agentic coding, with the largest gains on the hardest tasks. Anthropic reports that users hand off their most difficult coding work to Opus 4.7, work that used to need close supervision.
What you can do with it
Agentic coding
Opus 4.7 handles complex, long-running coding tasks with rigor and consistency. It pays close attention to instructions, finds ways to check its own output before reporting back, and keeps working on hard problems instead of giving up. On CursorBench, Opus 4.7 scores 70% against 58% for Opus 4.6. On Rakuten-SWE-Bench, it resolves three times as many production tasks as Opus 4.6.
Better vision
Opus 4.7 can read images at up to 2,576 pixels on the long edge, about 3.75 megapixels, more than three times the pixels of earlier Claude models. That lets computer-use agents read dense screenshots, pull data out of complex diagrams, and do work that needs pixel-exact references. On Replicate, max_image_resolution scales images down before sending them, to 0.5 megapixels by default and up to 2 megapixels.
Real-world work
The model makes better interfaces, slides, and documents for professional tasks. Anthropic reports that it leads the Finance Agent evaluation and GDPval-AA, a third-party evaluation of knowledge work in finance, law, and other fields.
Better instruction following
Opus 4.7 takes instructions literally instead of interpreting them loosely or skipping parts. Prompts written for earlier models may need adjusting.
Better memory across sessions
Opus 4.7 is better at using memory stored in files. It keeps important notes across long, multi-session work and uses them to start new tasks with less up-front context.
How it works
Claude Opus 4.7 has a one million token context window. It can write up to 128,000 tokens in one response, which you set with max_tokens (default 8,192).
The model decides when deeper reasoning would help and when to move quickly through simple parts of a task.
Improvements over Opus 4.6
- More rigor and consistency on long-running tasks
- Image input at three times the resolution
- Better interfaces, slides, and documents
- Stricter instruction following
- Better memory stored in files across sessions
- An updated tokenizer
When to use it
Use Claude Opus 4.7 for:
- Complex, autonomous coding work that runs for hours
- Analysis of high-resolution images, screenshots, or diagrams
- Multi-step tasks that need sustained focus
- Writing documents and presentations
- Financial analysis, legal reasoning, or research across many sources
- Long-running agent workflows with little supervision
When speed matters more than depth, try anthropic/claude-sonnet-4.6.
Notes
Opus 4.7’s updated tokenizer can turn the same input into 1.0 to 1.35 times as many tokens, depending on the content. It also thinks more on later turns in agent workflows. To use fewer tokens, lower max_tokens or ask the model to be more concise.
Anthropic reports low rates of concerning behavior such as deception, sycophancy, and cooperation with misuse.
Try Claude Opus 4.7 in the Replicate playground.