Anthropic may have just made life easier for developers.
Anthropic, a GenAI startup, has made new features available to developers and technical teams on its back-end console. These features claim to help create efficient prompts. Developers will be able to generate prompts, automatically create test cases, and compare outputs side by side with the new “Evaluate” tab.
Anthropic’s AI model Claude 3.5 Sonnet powers a built-in prompt generator that lets users describe tasks and have the AI large language model generate high-quality prompts for them. The company recently announced that this feature has now been made available to developers.
In order to test prompts to see how the LLM responds, users can either ask Claude to auto-generate test cases or manually add or import test cases that provide data for the variables. Users can then view generated test cases and modify them to suit their needs, which allows iterative improvements to the production process. This includes both generated and manual test cases.
Because users can quickly iterate by creating new versions of their prompt and running tests again, model evaluation and prompt refinement are now faster. Test cases from earlier prompts can also be easily generated again.
In order to assess performance, Anthropic has included the capability to compare the results of two or more prompts side by side. This allows for a better understanding of how a change or iteration on a prompt has affected the output.
The company has also stated that responses from AI models can be graded by subject matter experts using a five-point rating system to determine whether or not changes have improved the quality of the response.
“Prompt quality significantly impacts results,” Anthropic explained in a blog post. “But crafting high-quality prompts is challenging, requiring deep knowledge of your application’s needs and expertise with large language models.”
As of now, the Evaluation tool is in beta testing and accessible to all users on the Anthropic Console.
- Bing’s New AI Video Creator Uses OpenAI’s Sora – And It’s Free - June 3, 2025
- Opera Neon Redefines Web Browsing with Built-In AI Agents - May 28, 2025
- Google Workspace Gets AI Overhaul! - May 21, 2025
