A Data-Residency Constraint That Collapses the Decision Before Capability Even Matters

A prospective enterprise customer requires that any LLM feature must run entirely inside their own VPC, with no request ever leaving that network boundary — described in their security requirements as non-negotiable. A vendor proposes a well-known closed-API frontier model, arguing "it's simply the most capable model available, and capability should be the deciding factor."

1. Explain precisely why the VPC requirement rules out this proposal, regardless of how capable the model is.
2. What category of model does this constraint push the team toward, and why does that happen before any comparison of specific model capability scores is even relevant?
3. Suppose the team then also needs at least 100k tokens of context. Is that constraint as decisive as the VPC requirement? Explain the difference in how the two constraints should be weighed.