Summary of Key Points
Not long after Anthropic’s AI model Claude released version Fiber 5, developers accidentally discovered two additional internal testing models, which have not yet been officially announced. These models are codenamed “Mallow” and “Melon,” and only a few people currently have access to them. The official release of these models is still pending.
Detailed Analysis
1. What are these two “hidden models” exactly? (Although not officially announced, we can make some speculations)
These models are still in the testing phase, but we can infer some details from their names and the nature of the tests:
- Internal testing models are like new dishes in a restaurant that are first offered to regular customers for feedback—they are not intended for sale but to identify issues and refine the model’s functionality. For example, they might be testing new conversational styles, faster response times, or optimizations for specific tasks such as writing copy or solving math problems.
- The names “Mallow” and “Melon” suggest a more user-friendly interface. “Mallow” might be designed for more casual, friendly conversations, similar to talking to a friend, while “Melon” could be more efficient and direct, suitable for tasks that require quick action (such as organizing information). However, these are just speculations, and we will have to wait for official announcements to confirm the details. Clearly, Claude aims to bring new improvements to the existing Claude model.
2. Why are these models being tested in secret without an official release? (The company’s strategy)
There are three main reasons why the company has chosen to keep these models under the hood:
- Fear of failure: The new models are not yet fully polished, and any bugs (such as incorrect responses or lagging performance) could damage the company’s reputation if announced prematurely. By letting a small group of users test them first, the company can fix issues promptly before releasing the models to the public.
- Competitive advantage: The AI industry is highly competitive (with companies like OpenAI’s GPT-4o and Google’s Gemini). If Anthropic reveals new features too early, competitors could quickly copy them. Keeping the models secret helps maintain a competitive edge.
- Valuable feedback: Feedback from a small group of users (such as experienced developers or partner companies) is much more valuable than that from a large-scale public test. These users can provide specific feedback on areas that need improvement, helping to refine the models more effectively.
3. Can ordinary users use these models now? (Not yet, but it could be beneficial in the future)
Ordinary users cannot use these models yet, as they are only available for internal testing. However, this is good news for us:
- The strengths of these models will eventually be incorporated into the official versions. For example, if “Mallow” proves to be great for casual conversations, future versions of Claude may sound more natural. If “Melon” is fast and efficient, it could speed up tasks like information retrieval and writing.
- It’s also possible that these models are tailored for different user groups. “Mallow” could be designed for general users, while “Melon” is aimed at enterprises for more professional tasks, offering more choices in the future.
4. Is the AI industry becoming more competitive again? (Why is Claude rushing to release new models?)
The release of new models like “Mallow” and “Melon” reflects the intensifying competition in the AI industry:
- Major AI companies are competing to release the latest and most powerful models. OpenAI’s GPT-4o can process images and videos, and Google is working on Gemini. If Claude lags behind, users might switch to these alternatives.
- Both developers and companies are choosing the best models based on their performance and suitability for their needs. By releasing multiple internal testing models, Claude aims to quickly identify and address user pain points, thereby retaining its user base.
In summary, the emergence of these two models indicates that the competition in the AI industry is heating up. The AI tools available to us as ordinary users will continue to improve. For now, let’s wait patiently for the official release of these models and see what they have to offer.