Anthropic has launched Claude Opus 4.7, continuing its rapid iteration schedule with a model focused on complex, long-running tasks.
**What’s New in Opus 4.7**
The latest Claude model brings improvements across several key areas:
– **Better instruction following**: Previous Claude versions sometimes made assumptions or missed subtle instructions. Opus 4.7 is significantly better at following user requests exactly as specified.
– **Enhanced vision**: Support for higher resolution image recognition improves performance in UI, presentations, and document generation tasks.
– **Improved memory**: The new filesystem-based memory capability allows Claude to retain key information across sessions, reducing the need to repeat context.
– **Financial analysis**: Better performance in financial tasks makes Opus 4.7 more useful for business applications.
**Performance Comparison**
In benchmark tests, Opus 4.7 outperforms Claude Opus 4.6, GPT-5.4, and Gemini 3.1 Pro in several categories—though it falls slightly short of the more capable (but not yet widely available) Claude Mythos Preview.
Early testers including Intuit, GitHub, and Notion have provided positive feedback, particularly on the model’s ability to handle extended, complex tasks without losing coherence.
**Token Consumption Consideration**
One important note: Opus 4.7 uses a new tokenizer that results in approximately 1.0-1.35x more tokens for the same content. In high-compute mode, more thorough reasoning also generates additional output tokens. Users should factor this into their usage planning and budgets.
**Availability**
Claude Opus 4.7 is now available across all Claude products and APIs, including Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Pricing remains the same as Opus 4.6.
**The Anthropic Strategy**
With Opus 4.7, Anthropic continues to position Claude as the choice for enterprise users who value reliability and instruction adherence. The focus on long-running tasks and session memory suggests a strategy centered on professional productivity rather than casual use.
—
*Have you tried Claude Opus 4.7? Share your thoughts below.*
Leave a Reply