One of the biggest challenges faced by Tess AI users was choosing which text artificial intelligence model (LLM) to use for different tasks. With the LLM summary feature in the AI Copilot, it has become easier for you to choose which model is best for the goal you want to achieve. This article will guide you on how to use this new feature and better understand the characteristics of the available text AI models.
The LLM summary, or text AI model, can be found as a tag with an "i" next to the name of each model in the chat. In this tag, users can see a summary of the competencies and specialties of each LLM, making it easier to choose the right model for each task. This summary also includes an evaluation table that provides important details about the model's performance, as well as scores from 0 to 10 for each of the evaluated categories.
The AI model summary provides a detailed overview of the main characteristics and performance of each AI model. It includes essential information that helps in choosing the most appropriate LLM for the task you are performing.
Below are the elements that make up the summary:
A context window of an LLM refers to the maximum amount of text that the model can "remember" during a conversation or while processing information. It acts as the AI's memory, storing the context of the conversation, including what was said earlier.
The cost of the model is classified as low, medium, or high. This helps manage credit consumption when using the AI Copilot. Models with higher costs tend to be more powerful or robust, while cheaper models can be used for simpler or lower-impact tasks.
The speed of the model indicates the time it takes to process and respond. This speed can be fast, moderate, or slow, and your choice depends on the urgency and complexity of the task. Naturally, more robust and complex models tend to be slower, while simpler models tend to be quicker.
The Knowledge Cut-off Date informs the last update date of the model. This means that the model was trained with data up to this date. For example, if ChatGPT-4 was trained until October 2023, it will not have information about events or developments that occurred after that date.
For tasks requiring more recent and updated data, another option is to activate the internet query tool. Just select the internet option in the tools section, and the model will access the web to consider more up-to-date data and include it in the response.
In addition to the information in the table, the LLM summary also includes a performance evaluation in six categories, with scores ranging from 0 to 10. These scores help identify the model's areas of greatest competence.
The overall score reflects the total performance of the model across various areas and types of tasks. A model with a high overall score is versatile and can be used for a wide range of activities.
This category assesses the model's performance on questions related to natural sciences, such as biology, chemistry, and physics. Models with high scores in this area are ideal for scientific and academic tasks.
The score in programming measures how well the model can handle software development-related tasks, such as writing or reviewing code. A model with a high score in this area will be more efficient for those needing technical support in programming languages.
The common sense score evaluates the AI model's ability to handle everyday information and questions, providing answers based on general knowledge. Models with high scores in this area are great for tasks that require clear and accessible answers.
This criterion tests the model's skills in solving mathematical problems and performing calculations. If you are looking for an LLM that can handle numbers, spreadsheets, or complex formulas, a high score in mathematical analysis will be essential.
The reading comprehension score assesses the AI model's ability to interpret and understand texts. A model with a high score here will be able to process and answer questions based on text excerpts more effectively.
Consulting the summary before selecting a model is crucial to ensure a good use of your credits and optimize the results. By choosing the appropriate model, you can:
Better manage credit quantity: Each model has a different cost, and choosing one that best suits your task prevents wasting credits on models that may not be as efficient for what you need.
Adjust the context to the amount of material: Models with larger context windows can handle more information at once. This is essential when you are working with long or varied documents, ensuring that the model will process all the material without leaving anything out.
Improve answer quality: When you choose a model that is appropriate for the task at hand, the quality of the response improves significantly. This happens because the selected model is aligned with the nature of the task, bringing more precise and relevant results.
The LLM summary in the AI Copilot provides a clear and detailed view of the capabilities and limitations of each AI model. With this tool, you can optimize the use of the models, always choosing the most suitable one for each task, saving credits while ensuring the best quality in responses. Take advantage of this resource to make your interactions with the AI Copilot even more efficient!
