Why You Cannot Choose Your Own AI Model?
Some customers have asked whether they can select a different AI model for generating cover letters. The short answer is no, and here’s why that decision works in your favor.
Why We Standardize on One Model?
We regularly test multiple state-of-the-art AI models using real job descriptions and carefully designed guardrails. This process ensures that the chosen model:
Understands context correctly
Avoids mistakes like altering URLs
Produces professional and reliable results
After benchmarking, we select the best-performing model and apply our own tuning to further improve accuracy and consistency.
By keeping all customers on the same optimized model, we guarantee a consistent, high-quality experience. Allowing everyone to switch models individually would lead to uneven results and reduce the overall accuracy of your cover letters.
How the Numbers Back This Up (From Our Upwork AI job filtering)
We don’t expect you to just take our word for it, here’s what the data shows:
ROC AUC: 0.895, in nearly 9 out of 10 cases, the model correctly identifies and prioritizes strong job matches. That means the foundation for your cover letter is based on highly relevant opportunities.
F1 score: 0.75–0.78 (up from ~0.50), this reflects a massive 45–50% improvement in accuracy within a few months, leading to higher-quality cover letter drafts.
Customer satisfaction: mid-70s% (up from low-60s), users now see fewer irrelevant details and more polished results in their generated cover letters.
These improvements come directly from our process of standardizing on one optimized model and upgrading it when a clearly better option is available.
We’ve kept this article high-level, but if you’d like to dive into the charts and technical results, you can find them here: Upwork AI Job Filtering – Why GigRadar is the Best Agent.
The Bottom Line
We do not allow customers to change the AI model because doing so could reduce accuracy and lower the quality of your cover letters. Instead, we focus on continuously testing, upgrading, and tuning the best available models so you benefit from steady improvements without the risk of inconsistency.