Anthropic Unveils Opus 4.7: The AI That Outperforms Its Rivals

Anthropic Unveils Opus 4.7: An Advanced AI Model

Anthropic continues to accelerate in the field of artificial intelligence with the launch of Claude Opus 4.7. This model, designed for professional users, stands out for its improved programming performance and its ability to process high-resolution images. For several months, Anthropic has been rolling out versions of its Claude models to remain competitive against giants like OpenAI and Google. Just two months after the release of Opus 4.6, the company presents Opus 4.7 as its most powerful public model to date.

Remarkable Performance on Benchmarks

In its announcement, Anthropic highlights the measurable improvements of Opus 4.7 on benchmark tests. On the SWE-bench Pro test, which evaluates the performance of models in agentic programming, Opus 4.7 achieves a score of 64.3%, surpassing Opus 4.6, which scored 53.4%. This score also places Opus 4.7 ahead of competing models such as GPT-5.4 with 57.7% and Gemini 3.1 Pro with 54.2%.

On the SWE-bench Verified benchmark, Opus 4.7 achieves an impressive score of 87.6%, and it reaches 94.2% on GPQA Diamond, an advanced reasoning test at the doctoral level. Beyond raw performance, Anthropic emphasizes that the model has improved its ability to accurately follow instructions and verify its responses before providing them.

Designed for Long and Complex Tasks

Opus 4.7 is particularly suited for "agentic" uses, meaning systems capable of performing complex tasks with minimal human supervision. According to Anthropic, the model can handle multi-step projects over extended periods while maintaining coherence over a context window of up to one million tokens. This capability allows it to work on large projects, such as software development or document analysis.

The model's multimodal vision has also been enhanced. Opus 4.7 can now analyze images up to 2,576 pixels on the long edge, more than three times the resolution accepted by previous Claude models. This opens the door to uses such as analyzing complex screenshots or extracting data from detailed diagrams.

A Two-Tiered Strategy

Despite these advancements, Anthropic clarifies that Opus 4.7 is not its most powerful model. Internally, the company has a more advanced system called Mythos, currently reserved for a limited number of partners as part of the Project Glasswing cybersecurity program. This model reportedly exhibits even higher performance, particularly in tasks related to cybersecurity. However, Anthropic has chosen to limit its distribution to better control the risks associated with these capabilities.

In this strategy, Opus 4.7 thus plays an intermediary role: it allows for the deployment of certain innovations at scale while testing the necessary safeguards before potentially releasing even more powerful models. Available now via the Claude API, Amazon Bedrock, Vertex AI, or Microsoft Foundry, the model retains the same pricing as its predecessor: $5 per million tokens for input and $25 for output.