GameDev Assistant offers two AI model modes: a non-reasoning model (default) and a reasoning model (available only via
pro plans). The difference between these modes lies in the keyword of "reasoning", though both are generally capable of most tasks regardless of which you choose.
Non-Reasoning Model
Under the hood, non-reasoning models rely on pattern recognition, statistics, and machine learning to produce its results. The model isn't able to definitively explain how it arrived at the rendered results; rather it uses an immense pool of data to conclude which words are more likely to appear together given the context of the prompt.
From the user end experience, this model mode is the best general use case, which is why it is set as the default. Not only does it produce results very quickly, but it is less likely to result in errors caused by the LLM getting overwhelmed or timing out.
Reasoning Model
Reasoning models are somewhat just the opposite of non-reasoning models. These models were built to make logical conclusions by breaking down complex tasks, applying rules, and making inferences. Unlike their default counterpart, they can explain to you the exact logical steps they took to arrive at any one answer.
This model mode is a great choice if you need to do something extremely complex or need extensive research done into a topic. However, be aware that reasoning models take a significantly longer time to produce results, and have a larger generalized risk of failures with LLMs.