Run the agentsAvailable in v0.3.0
New in v0.3.0: benchmark scores pick the model. Difficulty and a profile — performance, balanced or efficient — set the score a job needs. Balanced takes the cheapest model that reaches it, efficient the best score per price, and performance the top scorer for the hardest work. No table ties a difficulty to a model.
A fixed table from difficulty to model goes stale every time a new model ships, and it cannot weigh price against quality.
[stages] names, the model is the first that applies: a model pinned for the tier, the first listed model in that tier's tier_pools, a quality_order naming a listed model, then the score rule.performance adds 1 and balanced adds nothing, and the step never goes below the floor of the task's kind.performance and balanced take the cheapest model by list price that reaches the step; efficient may go one step lower, never below the floor, and takes the most score per list price.balanced and efficient accept a model within 2 percent of the scale below the mark; performance and floors have no such margin. The tier still sets the effort.oah models and oah doctor show the profile, scores with source and date, each tier's model and each left-out model with its reason; oah recommend --engine <engine> --stage <stage> answers for one task, launches nothing and writes nothing.See the model choice for one task without launching anything (from Concepts).
oah recommend --engine <engine> --stage <stage>efficient can cost more than balanced at a low tier.OAH_MODEL_CATALOG_UNAVAILABLE) or no usable scores (OAH_MODEL_NO_SCORES), automatic choice stops; a pin gets past both, and oah never falls back to a default model.