Free tool
Model Picker
Which AI model should you use? Answer a few questions and get a transparent, explainable recommendation across Claude, GPT, and Gemini.
Get started-
Every recommendation traces back to a visible reason, not a black-box score — see exactly why a model was picked.
-
Weighs task type, latency sensitivity, and quality stakes together, not cost alone.
-
Shows the full cost comparison across every current model, across every provider — not just the recommendation.
-
Never picks a pricier model over a cheaper same-tier alternative without a cited reason — and never fabricates a single winner when two models are a close call (see the methodology page).
Frequently asked questions
It looks up your task type and quality stakes in a fixed, published table (not a black-box model) — each cell names the required capability tier and the reason. Latency and long-context needs then adjust the rationale you see, never hidden logic. See the full methodology page for the exact tables.
When the cost of a wrong or weak answer is higher than the price difference — customer-facing work, production coding, or anything with legal, financial, or medical stakes. For narrow, low-stakes tasks, the smallest model that does the job is usually the right call.
Every model is assigned one of four cited capability tiers (Small/Fast, Balanced/Mid, Flagship, Frontier-Extended) that are comparable across providers, not a provider-specific label. Within the tier your workload requires, the picker always recommends the current cheapest model — when two models are within about 5% of each other in cost, both are shown together rather than picking a fabricated winner.
The methodology page lists every model, its tier, and a source link for that assignment — nothing is a guess.