Anthropic Unveils Claude Opus 4.7: Try It, Safety Features, and Benchmarks
In a significant move, AI company Anthropic has launched Claude Opus 4.7, their latest model designed for general use. This new iteration aims to enhance user experience and tackle advanced tasks while balancing safety and reliability.
Claude Opus 4.7 Overview
Released in 2026, Claude Opus 4.7 is positioned as Anthropic’s most intelligent model available. However, it is noted that it does not surpass the capabilities of Claude Mythos. This model has been avoided for public release due to safety concerns.
Key Features of Claude Opus 4.7
- Hybrid reasoning models with multi-step reasoning and advanced coding capabilities.
- Improved in advanced coding tasks, visual intelligence, and document analysis.
- Higher quality in professional outputs like interfaces, slides, and documents.
How to Access Claude Opus 4.7
Users can access Claude Opus 4.7 through various platforms, including:
- Claude AI
- Claude API
- Anthropic partners like Microsoft Foundry
The model is priced similarly to its predecessor, Claude Opus 4.6. Due to its advanced processing capabilities, it utilizes more output tokens, which users should consider when managing their token limits.
Benchmark Performance
Benchmark testing has shown that Claude Opus 4.7 performs well compared to other models but falls short of the unreleased Claude Mythos. The following performance metrics were reported:
| Model | Humanity’s Last Exam Score (%) |
|---|---|
| Claude Mythos | 56.8 |
| Claude Opus 4.7 | 46.9 |
| Gemini 3.1 Pro | 44.4 |
| GPT-5-4 Pro | 42.7 |
| Claude Opus 4.6 | 40.0 |
When additional tools were used, GPT-5-4 Pro scored 58.7%, while Opus 4.7 achieved 54.7%. Mythos notably outperformed both with a score of 64.7%.
Safety Features and Improvements
Anthropic emphasizes that Claude Opus 4.7 maintains a low risk of misalignment, akin to its predecessor. It is less likely to hallucinate and shows reductions in reward hacking incidents.
The model card highlights:
- Increased reliability and honesty.
- Significant reductions in important omissions.
- Notable improvements in factual accuracy.
Overall, Anthropic’s Claude Opus 4.7 marks a significant step in AI development, offering users advanced functionalities while ensuring safety and reliability.
