Model Picker — describe your task, get a model recommendation

What do you want done?

The model behind this picker will classify the task, apply hard gates, and recommend the best fit. If your task has multiple stages (compress → generate → verify), the picker may return a multi-step plan.

Expected input tokens Expected output tokens

Try a quick example

Morning check-in: ask about sleep, training, supplements, mood Cron debug: bash script fails at 3am, no human review Deep research: compare 3 state management libraries for React, with citations Voice brief: summarize 3 articles for a 60-second voice read Architect: design a multi-tenant SaaS with isolation, billing, and observability

⚡ Keyword classifier (no LLM cost) — click to configure

This page asks an LLM to read the benchmark table and pick a model for your task, citing the numbers. It runs on the server — you don't need a key. The hard safety gates are enforced in code afterwards, so the LLM can advise but cannot talk its way past a non-hallucination floor.

If the recommender is rate-limited or unreachable, the page falls back to its weighted formula and tells you so. You can optionally paste an OpenRouter key below to have your key classify the task in that case; the ranking still comes from the formula.

OpenRouter API key (optional fallback)

Only used if the server recommender is unavailable. Stored in localStorage, sent only to openrouter.ai. Note: stored in localStorage in plaintext (not encrypted).