• Written by Azhar Aliyar, Director

In practice, the choice between open weight or proprietary models for AsscherAi, a platform that lets a large organisation query its own data in natural language and get answers in real time, often boils down to control and cost, with the operational burden each carries being a critical consideration. While some may argue that open weight models offer more flexibility, others claim that proprietary models provide better support and maintenance. However, the reality is that each option commits you to a specific set of constraints and requirements that can have a significant impact on your organisation's operations.

Commitments of each option

The open weight model commits you to maintaining and updating the model yourself, which can be a significant operational burden. This requires a team of skilled data scientists and engineers to ensure that the model remains accurate and effective over time. On the other hand, the proprietary model commits you to relying on the vendor for support and maintenance, which can limit your control over the model and its development. For example, a category manager comparing two suppliers using AsscherAi may find that the proprietary model's limitations hinder their ability to customise the model to their specific needs.

In contrast, the open weight model provides more flexibility and customisability, but at the cost of increased complexity and maintenance requirements. A shift supervisor checking the maintenance log using AsscherAi may appreciate the ability to tailor the model to their specific needs, but may also struggle with the added complexity of maintaining and updating the model.

The case for the other option

There are cases where the proprietary model is the right choice, despite its limitations. For example, if an organisation lacks the resources or expertise to maintain and update an open weight model, a proprietary model may be a more practical choice. In such cases, the vendor's support and maintenance can provide a level of stability and reliability that would be difficult to achieve with an open weight model. Additionally, proprietary models can provide a level of integration with other systems and tools that may be difficult to achieve with an open weight model.

On the other hand, there are also cases where the open weight model is the right choice, despite its added complexity. For example, if an organisation requires a high degree of customisability and flexibility in their model, an open weight model may be the better choice. In such cases, the ability to tailor the model to specific needs and requirements can provide a level of accuracy and effectiveness that would be difficult to achieve with a proprietary model.

Hidden costs

One of the costs that only shows up later is the cost of integration with other systems and tools. While proprietary models may provide a level of integration out of the box, open weight models often require significant customisation and integration work to get them working with other systems. This can be a significant cost, both in terms of time and resources. For example, integrating an open weight model with a legacy system may require significant custom coding and testing, which can be time-consuming and expensive.

Another cost that can show up later is the cost of maintenance and updates. While proprietary models may provide a level of support and maintenance, open weight models require ongoing maintenance and updates to ensure that they remain accurate and effective over time. This can be a significant cost, both in terms of time and resources. To learn more about how AsscherAi can help with maintenance and updates, visit our website.

Deciding without a six month evaluation

One way to decide between open weight and proprietary models without a six month evaluation is to conduct a thorough analysis of your organisation's needs and requirements. This includes considering factors such as the level of customisability and flexibility required, the resources and expertise available for maintenance and updates, and the level of integration required with other systems and tools. By carefully considering these factors, you can make an informed decision about which option is best for your organisation.

Another approach is to start with a small pilot project to test out the open weight or proprietary model. This can provide a level of feedback and insight that can inform your decision, without requiring a full-scale implementation. To get in touch with our team and discuss your options, visit our contact page.

When it's the wrong choice

There are cases where neither the open weight nor proprietary model is the right choice. For example, if an organisation requires a high degree of accuracy and effectiveness, but lacks the resources or expertise to maintain and update an open weight model, and is also unable to afford the cost of a proprietary model, then neither option may be viable. In such cases, it may be necessary to consider alternative approaches, such as developing a custom model from scratch or using a different type of system or tool altogether.

In general, the choice between open weight and proprietary models requires careful consideration of a range of factors, including control, cost, and operational burden. By carefully evaluating these factors and considering the specific needs and requirements of your organisation, you can make an informed decision about which option is best for you. Most people get wrong the idea that open weight models are always more flexible and proprietary models are always more expensive, but this is not always the case, and a thorough analysis of the specific situation is required to make the right choice.