Anthropic has released Claude Opus 4.7, a significant upgrade to its flagship AI model. The new version offers better coding capabilities, sharper vision, and a novel ability to double-check its own work. However, the company publicly acknowledged that this release does not match the performance of its unreleased, more advanced system called Mythos.
This concession highlights the widening performance gap between publicly available AI models and the most advanced systems being developed in private. Anthropic cited safety concerns as the reason for withholding Mythos from general release. The firm has only provided access to a select group of technology and cybersecurity companies for evaluation.
According to the company's announcement, Opus 4.7 outperforms its predecessor, Opus 4.6, as well as competing models like ChatGPT 5.4 and Google Gemini 3.1 Pro in several key benchmarks. Anthropic stated the model shows particular gains in advanced software engineering, especially on the most difficult tasks. Users reportedly can now hand off their hardest coding work to the model with confidence.
The release underscores the strategic tension between rapid commercial deployment and cautious safety protocols within the AI industry. By openly benchmarking its public offering against a superior internal system, Anthropic is signaling both its technical progress and its commitment to responsible development. This move may pressure competitors to be more transparent about their own internal capabilities.
Industry observers note that such disclosures could become a new norm, serving as a form of technical marketing while managing public expectations about the pace of advancement.