Choosing Modelsintermediate12 min · 9 steps
Choosing a Model
A practical decision path from constraints to a model — then cascades, cross-vendor fallbacks and an upgrade pipeline so the choice survives monthly releases.
requestresponseheavy workerror
ProblemStep 1 of 9
No model wins on every axis
Leaderboards rank one dimension. Production cares about seven: agentic reliability, cost per task, context recall and long-context pricing, rate-limit ceiling (and whether a spend cap can hard-stop you), data handling (ZDR, residency), ecosystem maturity and lifecycle runway.
Start from your binding constraint — ceiling, open weights, agents, prompt length, volume — not from whoever topped the latest index.
Watch outAnthropic’s self-serve tiers have hard monthly spend caps; hitting one returns 429 with no retry-after until the 1st of next month. A budget cap is an availability risk.
Lesson 4 of 4 in Choosing Models