OmnisBench vs RouterBench

A citable academic baseline, and a live one you can re-run today.

RouterBench (Martian, arXiv:2403.12031) is a respected, peer-reviewed academic benchmark for LLM routing — a frozen dataset and paper. OmnisBench takes the same core idea and ships it as a live, one-command-reproducible tool: current models, pinned current prices, and results anyone can independently re-grade offline. Apache-2.0.

At a glance

How they compare

DimensionRouterBenchOmnisBench
TypeAcademic paper + frozen datasetLive, open-source CLI tool
Models coveredFixed set, pinned at publication (2024)Config-driven pool — currently Claude Opus 5, GPT-5, Claude Haiku 4.5, GPT-5-nano
FreshnessFrozen at publication datePinned pricing snapshot, re-runnable against current models anytime
Reproducible todayDataset available; not designed to re-run against today's modelsYes — omnisbench run against any configured pool
Independently verifiablePeer-reviewed methodology, published paperYes — omnisbench verify re-grades published responses offline, zero API calls
LicenseAcademic dataset releaseApache-2.0
Cost / pricing basisPrices as of 2024 publicationPinned, dated pricing snapshot (2026-08-18), reproducible as list prices move
Freshness & current models

The model landscape moves faster than a paper can

RouterBench did valuable work pinning down a methodology for scoring routing policies against a quality-cost frontier. But its model list and prices are frozen at publication — new frontier and budget models simply aren't in it. OmnisBench's model pool and pricing snapshot are config, not code, so a new run against today's models (or tomorrow's) is a config change and a re-run, not a new paper.

Reproducibility & verification

Two different kinds of trust

RouterBench earns trust through peer review and a citable, fixed methodology — the right bar for an academic result. OmnisBench earns trust differently: every model response for every task is published in results.json, and omnisbench verify re-runs the graders against those responses and re-derives the whole leaderboard offline, with zero API calls. Change one stored answer and it fails. That's a complementary, tool-shaped kind of verifiability — not a replacement for peer review, but a way to check a number yourself in minutes.

Methodology lineage

OmnisBench builds on ideas RouterBench helped establish

We say this plainly: RouterBench helped establish the idea of scoring LLM routing policies against a quality-per-dollar frontier, and that framing is core to how OmnisBench presents its own leaderboard. OmnisBench isn't claiming to out-rank RouterBench's rigor — it's a different artifact (a live tool, not a paper) built for a different job: getting a fresh, checkable number against whatever models you actually care about today.

Who each is for

Pick based on the job

RouterBench

You want a citable, peer-reviewed methodology and baseline for academic work, a paper to reference, or a fixed historical comparison point.

OmnisBench

You want to know, right now, how much routing efficiency is actually available across today's models and prices — and you want the number to be independently checkable, not just asserted.

See the numbers

OmnisBench's first run

Ideal routing hit 99.7% task success at ~90% lower cost than always using the frontier model — and every figure is re-gradable offline.

See the full leaderboard on the homepage →
Reproduce it on GitHub →
Read the open-source LLM routing benchmark overview →

Questions

Frequently asked

Is OmnisBench a RouterBench alternative?
OmnisBench is not trying to replace RouterBench so much as build on what it established. RouterBench is a respected, citable academic dataset and paper; OmnisBench is a live tool you run against today's models and prices, with results independently re-gradable offline. Use RouterBench for methodology and citation, OmnisBench for a current, verifiable number.
Why is RouterBench considered frozen?
RouterBench (Martian, arXiv:2403.12031) is a fixed dataset and paper published at a point in time. Its model list and prices reflect what was available then, and re-running it against today's models and pricing isn't part of its design. That's normal for an academic baseline, but it means the numbers age.
What does OmnisBench add on top of RouterBench's ideas?
OmnisBench keeps the core idea RouterBench helped establish — score routing policies against a quality-per-dollar frontier — but ships it as a live, Apache-2.0 command-line tool with a pinned pricing snapshot, so you can re-run it against current models and independently verify the published results offline with zero API calls.