Anthropic Unveils Claude Opus 4.7 With Stronger Coding and Vision

• Anthropic introduced Claude Opus 4.7 as its most capable Opus model yet. 
• Anthropic says Opus 4.7 improves coding, multimodal understanding, and sustained reasoning over longer tasks. 
• Pricing remains unchanged at $5 per million input tokens and $25 per million output tokens.

Anthropic has launched Claude Opus 4.7, presenting it as the company’s most capable Opus model so far and a more reliable option for difficult, long-duration tasks. In its launch message on X, Claude said the new model handles long-running work with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. 

That framing matters. Anthropic is leaning less on raw spectacle and more on dependability. The company’s official release says Opus 4.7 is designed to investigate more effectively, sustain reasoning over long runs, and improve performance in real-world engineering workflows. 

Anthropic Focuses on Coding and Long-Task Reliability

Anthropic’s official materials and early tester comments point to software engineering as one of the clearest upgrade areas. The company highlights gains in code review, debugging, terminal work, planning, and multi-step execution, while outside testers cited stronger autonomy, lower tool error rates, and better follow-through on difficult tasks. 

Source: X

The benchmark table shared alongside the launch shows Opus 4.7 posting stronger results than Opus 4.6 across several coding and reasoning categories, including agentic coding on SWE-bench Pro, scaled tool use, agentic computer use, and visual reasoning. In the chart, Opus 4.7 also compares competitively with other frontier systems across graduate-level reasoning and multilingual Q&A. 

Additionally, Anthropic is clearly positioning Opus 4.7 as a production-grade step rather than a pure frontier demo. The company says the model is better at checking its own work and staying coherent over longer workflows, which directly addresses one of the biggest complaints developers have had with agent-style systems. 

Vision Support and Safety Moves Higher

Anthropic also says Opus 4.7 improves multimodal understanding, especially for dense screenshots, technical diagrams, and data-heavy visual material. Reporting on the release says the model now supports images up to 2,576 pixels on the long edge, giving it a clearer edge in detailed image analysis and computer-use tasks. 

However, the company is still drawing a line between Opus 4.7 and its restricted Mythos Preview system. Anthropic says Opus 4.7 is less broadly capable than Mythos Preview, even though it improves on Opus 4.6 across a wide range of benchmarks. 

Security remains central to the rollout as well. Anthropic says it has built in safeguards that automatically detect and block prohibited or high-risk cybersecurity uses while also launching a Cyber Verification Program for vetted security professionals. 

Availability Stays Broad, and Pricing Stays Flat

Opus 4.7 is available through Anthropic’s API and across major cloud platforms, including Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. GitHub also said the model is generally available in Copilot for Pro+, Business, and Enterprise users. 

Notably, Anthropic did not raise prices with the upgrade. Multiple reports on the launch say pricing remains at $5 per million input tokens and $25 per million output tokens, keeping the commercial terms in line with the prior model. 

The release shows where Anthropic wants to compete next. Rather than chasing attention with a bigger promise alone, it is pushing Claude Opus 4.7 as a steadier model for real engineering, better image reasoning, and lower-supervision work in production.

Related: Drift Protocol Plans $147.5M Recovery After Major Exploit Hit