Anthropic has officially rolled out Claude Opus 4.7 , presenting it as a significant step up from its predecessor, Opus 4.6. The new model is designed to deliver stronger performance in areas such as advanced software development, complex multi-step reasoning, and high-level professional tasks that require deeper knowledge and precision.
According to the company, Claude Opus 4.7 is now accessible across its ecosystem, including Claude applications, its API, and major cloud platforms like Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Despite the upgrades, pricing remains the same as the previous version, set at $5 per million input tokens and $25 per million output tokens.
Read Also: Anthropic Usage Bundles Explained: Are You Overpaying For Claude Third-Party Tools?
Anthropic highlighted several technical improvements in this release. The model is better at following instructions accurately, can sustain longer and more demanding tasks with improved consistency, and offers enhanced vision capabilities. Notably, it can now process high-resolution images up to 2,576 pixels on the longest side.
Internal evaluations show that Opus 4.7 outperforms version 4.6 across a range of use cases, including coding, financial analysis, document processing, and agent-like workflows. The update also introduces a new “xhigh effort” mode, which allows users to balance deeper reasoning with response speed. However, Anthropic cautioned that users may notice differences in behavior, as the model tends to interpret instructions more literally than earlier Claude versions.
This release comes shortly after the debut of Claude Mythos Preview, introduced on April 7 through Project Glasswing. Anthropic described Mythos as its most powerful model so far, particularly noting its strong capabilities in cybersecurity applications. The initiative is aimed at improving the safety and resilience of critical software systems.
For now, Mythos Preview remains restricted, while Opus 4.7 becomes the first model to incorporate new safety mechanisms designed to identify and block risky or prohibited cyber-related requests ahead of any broader Mythos rollout.
The launch of Opus 4.7 also coincides with intensified competition in the AI space. OpenAI recently introduced GPT-5.4, positioning it as a highly capable and efficient model for professional tasks, with advanced performance in coding, tool usage, and long-context processing. It also followed up with lighter variants, GPT-5.4 mini and nano, aimed at faster and more specialized workloads.
Meanwhile, Google has continued to expand its Gemini lineup at a rapid pace. After launching Gemini 3.1 Pro in February for handling complex operations, the company introduced Gemini 3.1 Flash Lite and Flash Live in March. Most recently, on April 15, it unveiled Gemini 3.1 Flash TTS Preview, further pushing into low-latency and multimodal AI capabilities.