Nadir runs beside your AI gateway, starts in shadow mode, and picks the cheapest model, reasoning effort, cache, and context plan that still fits each agent task. Attach measured usage and outcomes to every decision receipt, then enforce only after the evidence clears your quality bar.
Nadir is a verifier-gated cascade router for LLM traffic. A trained pre-classifier scores each prompt in under 10 ms and picks the lowest viable tier, the cheap model answers, and a verifier scores that answer. Escalation to a stronger model happens only when the answer misses the configured bar. Nadir is OpenAI compatible, so migration is a two-line change: point the base URL at api.getnadir.com and set model to auto. Bring your own keys is supported on every tier.
Nadir can also run beside an existing gateway as a decision-only API, with no provider call and no response traffic, which is how a shadow-mode evaluation works.
On a reference-assisted RouterBench evaluation over 11,420 held-out triples, the cascade ran at 60% lower cost than always-Opus while retaining 98% of its quality, with a 1.7% catastrophic-route rate. Separately, Nadir-Tumbler posts an arena_score of 72.3 on RouterArena's public scorer, 5th of 23 routers, which measures the routing decision on RouterArena's own model pool.
Those are research results. They are not a production guarantee, a universal savings rate, or a forecast for any particular workload. Customer savings are reported from measured execution against a declared baseline, and customer quality only from outcome-labelled traffic. Projected savings and realized savings are separate artifacts and are never blended.
Three rungs, picked by risk appetite. Rung 0 Shadow runs advisory decision calls alongside live traffic and returns a projected receipt, with nothing in the request path changed. Rung 1 Hosted is the two-line swap on a production slice and returns a realized receipt. Rung 2 On-prem is a supervised six-week proof of concept inside the partner's VPC, where no prompt, response, or usage reaches Nadir. The commitments are the same at every rung.
NadirClaw is the self-hosted core, source-available under the PolyForm Noncommercial License. Source-available is the correct label; NadirClaw is not open source. Nadir Route's hosted plan has no base fee and charges a variable fee only on measured savings from requests Nadir executed.
Machine-readable summaries of this site: llms.txt and llms-full.txt.