Ramp Router pricing and routing, compared with Nadir
Ramp Router (router.com) is free through 2026 at list token price. How it picks models, where it beats Nadir, and how Nadir differs, cited from Ramp's docs.
Questions answered on this page
What is the difference between Nadir and Ramp Router?
Ramp Router gives you one endpoint and free routing through 2026. Nadir decides which model each prompt needs, and can decide it without taking the prompt. Ramp Router is Ramp's hosted LLM gateway at router.com: one OpenAI- and Anthropic-compatible endpoint and one key for US-hosted models from OpenAI, Anthropic, Google, xAI and open-weight providers, with Flex-tier cost routing on by default. Free routing through 2026 with $26 in credits, running 2.75T+ of Ramp's own tokens a month.
How does Ramp Router decide which model to use?
By default the caller names the model, optionally pinning provider and service tier, and Router adds in-order fallback. Cost-efficient routing, on by default, keeps the model and moves eligible calls to the cheaper, slower Flex tier when recent latency allows. Benchmark aliases sample each request from a benchmark-weighted model mix. Switchyard, NVIDIA's open-source NeMo stage router, moves coding-agent turns between a frontier and an efficient model on tool-output errors, compaction, passing tests and turn depth. Jev Auto Routing, in limited availability, reroutes only when the latest turn clearly fits another listed model. Nadir decides from a trained difficulty classifier over the prompt, the candidate menu, policy, cache, context, and session economics.
What does Ramp Router do after the model answers?
Nothing today. Shadow models mirror a percentage of traffic to up to three models and record cost and latency, but the LLM-judge agreement metric is marked coming soon and shadow results do not affect routing. Nadir's decision-only mode returns before generation. On the managed, non-streaming path a complete response can enter a reference-free verifier and escalate when it misses the configured bar. Streaming bypasses that post-generation cascade.
What does Ramp Router cost compared with Nadir?
Ramp Router: Free through 2026, list price on the tokens you use, and the first $26 in credits included. No Router charge on your own provider keys. Pricing after 2026 is unannounced. US developers and teams only for now, with enterprise features listed as coming soon. Nadir has no base fee. Credits for Nadir-managed keys draw down at provider list price, with a 5.5% platform fee (8% on Business) added when you buy them. On your own keys, your provider bills you directly and Nadir adds no fee. Routing savings carry no fee, and shadow and recommendation-only deltas are advisory estimates that are never billed.
What does Ramp Router do well?
A contamination-free internal coding benchmark, Ramp SWE-Bench. Free routing through 2026 at provider list price. Flex-tier savings, and Switchyard turn-level switching for coding agents.
Why pick Nadir over Ramp Router?
Their default path asks the application to name the model. Prompt-aware choice is Jev Auto Routing, still limited, matching the latest turn to model descriptions you write, with effort pinned per model. Nadir returns a per-prompt model and effort decision. Their own cost chart pairs the 40% with a Flex share climbing from 1% to 73%, so most of that saving is a slower provider tier rather than a better model choice. Nadir measures savings against a baseline the customer declares, and takes no fee on them. Router records inputs, outputs and tool calls by default for a year unless you opt out, and it is US-only with enterprise still pending. Nadir can run decision-only with no content leaving your stack, self-hosted or whitelabelled, inside or outside the US. Nadir publishes no universal deployed savings or quality percentage. The 60% cost and 98% quality-preserved RouterBench result, measured on 11,420 held-out triples, is a reference-assisted research ceiling rather than the live reference-free path.
How do I trial Nadir against Ramp Router?
Keep Ramp Router exactly as it is. Write down the model, reasoning effort and monthly spend it produces today, and pick a week of real traffic, or a set of requests you can grade, as the baseline. Call Nadir's decision API from the same place that calls Ramp Router, in shadow mode. It returns the model, effort, cache and policy recommendation with no provider call, so nothing about execution changes. Read the receipts against what actually ran, and report outcomes through the outcome endpoint. Projected savings stay advisory until you execute a recommendation and report measured usage. Enforce on one workload, keep Ramp Router as the route you fall back to, and widen only when the outcomes clear your own bar.