Four-question check

Start from the API. Test for exceptions.

No tokens/day, GPU, quantization, or cloud choices. RunWhere only asks for what a cost-anxious API user already knows.

1. Current monthly API spend

Less than $200/month $200–$2K/month $2K–$10K/month $10K+/month

2. Hosted model in use

Hosted API model or family

3. Traffic shape

Bursty Peaks are at least ~5× average; scale-to-zero can matter. Steady Traffic stays within roughly ~2× average. Predictable batches Offline jobs, backfills, or scheduled inference.

4. Hard constraints

None Data residency Air-gap Latency <100ms p99

I know my candidate open-weight model

Candidate model (optional)

Boundary artifact runwhere-boundaries--v1.1.0-93a0a727