DecisionBench and choosing a decision model
Leaderboards are starters, not finishers
DecisionBench-style leaderboards (community posts track pplx-decider checkpoints topping accuracy-per-cost charts) are useful priors. But your distribution is not the benchmark’s: your states, your criteria, your ambiguity profile.
Run your own
The Playground takes your state and questions and runs them against any model in the catalog, showing latency, tokens and per-run cost. Paste a sample of your real traffic, try the top three models, and let your own numbers pick.
Guides & comparisons
- Semantic Router alternative: decision models for routing decisions
- RouteLLM vs decision models: routing between models vs deciding directly
- Decision models on OpenRouter vs a dedicated decision gateway
- Run pplx-decider locally with Ollama, or call it hosted?
- Decision model pricing comparison — the full catalog, live
- typesafe-sdk quickstart: swap base_url, keep everything else
- System 1 vs System 2 models: decision models vs generative LLMs