"No Vendor Lock-In" Is Code for "No Product"

The debate over "no vendor lock-in" in AI product design is a hot topic, particularly in relation to using multiple models for a single use case. Critics argue that this approach is simply a poorly thought-out workaround for a product that is essentially a clone of a successful one offered by frontier labs. However, proponents of the model picker argue that it allows users to access the very best models available today, such as GPT-6 Astra, Claude Fable 5.1, and Opus 5.5, which are significantly better than open-source alternatives.

The key issue is that users want the best performance for their specific task, not just the cheapest model. Open-source models, while cheaper and improving, are still a long way from matching the capabilities of frontier models. In practice, users should have access to the best possible model for their needs, without having to compromise on quality by switching to a cheaper, less effective model.

The focus should be on building a product that delivers the best user experience for a given task, regardless of the underlying model used. Companies like OpenAI and Anthropic are recognizing this and designing their products to provide the best possible user experience, regardless of which model is selected underneath.