A smaller classic model: Which model choice best fits these constraints?
A smaller classic model best satisfies explainability, on-premise hosting, and limited compute for an auditable regulated decision.
The question
A regulated lender needs a model for an auditable, explainable credit decision. It must run on-premise on modest hardware, and every decision must be defensible to auditors. Which model choice best fits these constraints?
Preparing for AIGP? Take the free 5-min readiness quiz →
- A large proprietary generative model accessed via API, which offers state-of-the-art fluency and offloads infrastructure and maintenance burdens onto the vendor.Plausible for capability, but an external API violates the on-premise constraint and generative opacity undermines auditability.
- A large open-source generative model fine-tuned in-house, giving full weight access and customization while still delivering broad general-purpose language capabilities.Plausible because it is on-premise, but its size strains modest hardware and its probabilistic reasoning is hard to explain to auditors.
- A multimodal generative model, which handles text and images together and future-proofs the deployment against later expansion of the credit use case.Wrong because multimodal capability is unnecessary here and adds cost and opacity that conflict with the explainability and compute constraints.
- A smaller classic (non-generative) model, which is typically more interpretable, cheaper to run on-premise, and better suited to a regulated, auditable decision. ✓Correct because a compact classic model aligns with the explainability, on-premise, and compute constraints the scenario imposes.
The trap
Assuming a larger or generative model is always the stronger choice, ignoring explainability and compute constraints. How to remember it
A smaller classic model best satisfies explainability, on-premise hosting, and limited compute for an auditable regulated decision.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Deployment and Use questions
- Data representativeness and workforce readiness gaps are: In assessing the use-case context before deploying, →
- Retrieval augmented generation: Which deployment technique augments a model's responses with information →
- To systematically identify and evaluate the potential: What is the primary purpose of performing or reviewing →
- All 424 Understanding How to Govern AI Deployment and Use questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use ·
Every answer, right and wrong, comes with its own explanation.
