NeoAI / Routers
Model routers pick one model for your prompt. NeoAI delegates every step inside it.
Per-request model routers like Martian and RouteLLM predict whether an entire prompt is hard or easy, then route the full request to either a small or frontier model. NeoAI operates at the sub-step level: it decomposes tasks into discrete steps so routine work runs on small models while reserving frontier reasoning for complex sub-steps.
Last updated 25 Oct 2026 · By NeoAI
The problem with all-or-nothing routing.
Per-Request Router
Routes 100% of tokens to frontier pricing if any step needs reasoning.
NeoAI Step Delegation
Runs routine extraction on small models; only reasoning reaches frontier.
Economic Impact
Cuts total spend 60–80% on mixed tasks without dropping quality.
Worked example on published list prices. Illustrative Price data
Side-by-side technical comparison.
| Capability | Per-Request Router | NeoAI Delegation |
|---|---|---|
| Decision Granularity | 1 Decision per prompt call | 1 Decision per task sub-step |
| Model Mixing | Single model handles entire prompt | Mixes Small, Mid, Frontier within DAG |
| Classification Method | Embedding classifiers or LLM judge | Heuristic planner & learned DAG rules |
| Human Rework Reduction | None (blind to post-generation edits) | Distils reviewer redlines into gates |
| Failure Handling | Re-routes failed request or throws | Escalates individual step one tier up |
| Knowledge Transfer | Zero transfer across workflows | Reusable recipe blocks across related tasks |
Comparison reflects published router capabilities (Martian, RouteLLM) as of October 2026.
Routers forget every call. Recipes compound know-how.
Learning carries over between related workflows
When a team moves from CV screening to performance review, 4 of 5 blocks are reused: rubric evaluation, evidence extraction, citation validation, and summary drafting.
Unrelated workflows start cold. Related workflows start ahead. Illustrative
Reused Workflow Blocks (Performance Review):
Questions buyers ask before they book.
Why isn't a per-request router enough for complex workflows?
Most enterprise workflows are mixed: a 10-page document analysis has simple data extraction (cheap) and difficult legal reasoning (expensive). A per-request router must classify the whole task as hard and route 100% of the tokens to a frontier model. NeoAI decomposes the task so each sub-step runs on the cheapest sufficient tier.
Does NeoAI require training router classifiers?
No. Model routers often require fine-tuning custom routing classifiers or training embedding models on historical data. NeoAI uses explicit, interpretable task recipes with deterministic quality gates (schema tests, citation checks, score bounds) distilled directly from your reviewers' accepted work.
How does recipe learning carry over between tasks?
Model routers treat every prompt in isolation. NeoAI recipes are constructed from modular blocks (rubric scoring, evidence extraction, calibration). When you launch a related task, it reuses the proven blocks and begins with higher quality from day one.
See your own numbers in about 2 weeks, free.
A free audit maps where your AI spend and rework go and sizes the savings on your data. No measured savings, no fee.