Overview › Lever

Model

Model choice, routing, cascades, reasoning effort. 5 optimizations change this part of an agent.

1. 1.b.i

   Match the model and its effort to each call site and query class

   Pick the model per role or call site rather than one model everywhere, and set its reasoning effort and sampling parameters per site; tune each role's prompt to its model (GEPA raised every model tested at lower cost).

   Models › Finding specialists

   Better qualityLower cost

   Tier 3 · tuning · changes model, prompt · any agent · 19 sources, 6 measured

2. 2.a.iii

   Choose how the agent acts: typed tools, code execution, or shell, matched to the model and task

   Decide the action interface deliberately, by model and by task, and provision what it needs.

   Tools › Refactoring

   Lower costMixed quality

   Tier 3 · tuning · changes tool surface, model · tool use agents · 22 sources, 10 measured

3. 3.c.i

   Self-Discover reasoning structures

   Self-Discover has the model compose a reasoning structure once per task type (select useful modules from 39 seed heuristics, adapt them to the task, write them as a step-by-step plan), then follow that cached structure on every instance, so each instance costs one call.

   Prompts › Sizing reasoning effort

   Mixed qualityHigher costSlower

   Tier 3 · tuning · changes model, prompt · reasoning agents · 11 sources, 9 measured

4. 6.a.i

   Oracle-worker multi-model

   Let a cheap model do the work and call an expensive one only when a trigger fires: a cascade escalates when the cheap answer looks unreliable or fails a check, a router picks the model before the call, and an advisor tool lets a cheap agent consult a stronger model mid-task.

   Topologies › Escalating

   Better governanceMixed qualityMixed costMixed performance

   Tier 4 · restructuring · changes model · any agent · 6 sources, 5 measured

5. 9.c.iii

   Audit a verifier-gated cascade against a gold sample the verifier never sees

   When a cheap model's outputs are accepted or escalated by a verifier, measure true error on a small gold-labeled sample outside the verifier: metrics measured through the verifier showed about 3% error while true error was 14-32%.

   Verification › Sizing verification

   Better qualityHigher cost

   Tier 4 · restructuring · changes control flow, model · any agent · 13 sources, 13 measured