# Undominated.ai — the AI inference price index > Ranks AI language models by independently measured capability first, then > compares what they actually cost. Quality leads; price breaks ties. Undominated.ai exists because the spread between the cheapest and most expensive model is about 3,553x, and most rated, priced models have another on the same page that scores at least as high, costs no more, and wins on one of the two, though not always a like-for-like replacement. 133 of 141 rated, priced models have an alternative with a higher score or a lower price, with neither dimension worse. 8 are not dominated on those two axes. **Citable facts and interpretation rules** Published prices and independent scores are source-linked. Derived ranks, workload prices, break-evens, and hardware-fit figures name their assumptions. They are not vendor list prices. - The published capability score is LMArena Elo, used under CC BY 4.0 with attribution. We never score models ourselves and never blend evaluators into one number. Artificial Analysis figures are not published here: we hold no redistribution licence, so they inform what we check and never what you read. - Ranks are significance ranks. A model's rank is one more than the number of models that beat it by more than both margins of error combined. Models with the same count share a rank. Two models within error of each other can still hold different ranks, because each is compared with the whole board. 141 ranked positions collapse to 74 genuinely distinct ranks. - Dominated means another rated, priced model with the same delivery mode scores at least as high, costs no more, and is strictly better on one of the two. Scores are compared as published Elo, not significance ranks. We only call the dominating model a clean replacement when it keeps every recorded capability: context window, maximum output, input modalities, tool use and reasoning. - 45 models change price past a context threshold, so a headline rate is the wrong number for long-context work. - 61% of the catalogue's models (216 of 357) have no score on the LMArena general board this site ranks on. Those models are listed as unrated, never ranked and never scored zero. **Licensing** LMArena leaderboard data is used under CC BY 4.0 from the official dataset. Artificial Analysis figures are not published on this site: we hold no redistribution licence, and we grant no sublicence to redistribute them. Clean Markdown is available by appending `?format=md` to a page URL. `Accept:` is not a route to it: Cloudflare HIT-caches the HTML for the canonical URL, so the origin never sees the header. Agents must never infer a price, score, or rank that is not in the cited data; unrated is not zero. Every model page carries an embeddable verdict badge at `https://undominated.ai/badge/.svg`, where `` is the segment that page already uses at `/models//`. A badge states one model's dominance verdict at the balanced workload on the LMArena lens, with the date it was computed, and it changes when that verdict changes. Standard-delivery, non-retired models have one; batch and free variants do not. Corrections supported by a primary source are applied directly under the published correction policy. ## Core pages - [Leaderboard — Markdown](https://undominated.ai/?format=md): 357 models, 141 ranked capability-first. - [Report a wrong price](https://undominated.ai/report/): Public GitHub issue: model, observed value, proposed value, primary source, date. - [Value frontier — Markdown](https://undominated.ai/frontier/?format=md): Models nothing dominates on capability and price. - [Plans — Markdown](https://undominated.ai/plans/?format=md): AI subscription plans and API break-even points. - [AI Tools — Markdown](https://undominated.ai/tools/?format=md): Compare AI coding tools and model clients by workflow, model access and platform. - [Skills — Markdown](https://undominated.ai/skills/?format=md): Curated skills with source reviews, selection reasons, limitations and sourced installation guidance. - [Agents — Markdown](https://undominated.ai/agents/?format=md): Reviewed agent role definitions with host requirements and licensed, attributed originals. - [MCP Servers — Markdown](https://undominated.ai/mcp-servers/?format=md): Reviewed MCP servers by task, language and transport, with permissions and setup evidence. Skills, agents and MCP servers are editorial selections based on the cited source material. A source review does not establish runtime reliability, security or comparative performance. Each review states its date, inspected material, findings and untested scope. - [Methodology and corrections — Markdown](https://undominated.ai/methodology/?format=md): Ranking method, limitations, provenance, and correction policy. - [Independence — Markdown](https://undominated.ai/independence/?format=md): Funding, commercial boundaries, and evidence commitments. - [Self-host — Markdown](https://undominated.ai/self-host/?format=md): Hardware-fit calculator for open-weight catalogue models. - [Open weights — Markdown](https://undominated.ai/open-weights/?format=md): Catalogue rows with openWeights=true. Null is not open. - [Batch — Markdown](https://undominated.ai/batch/?format=md): Standard vs batch twins. Same model, later delivery. - [Cached input — Markdown](https://undominated.ai/cache/?format=md): Published cached-input rates. Equal to fresh is not a discount. - [Now — Markdown](https://undominated.ai/now/?format=md): Dated snapshot of the current value frontier. - [Check a model — Markdown](https://undominated.ai/check/?format=md): Test whether a model is still undominated under the selected workload. - [Audit a bill — Markdown](https://undominated.ai/audit/?format=md): Client-side usage-export audit: strict dominance at the observed mix. - [Repo Check — Markdown](https://undominated.ai/check-repo/?format=md): Client-side README or package.json model-slug scan. Never paste .env. Browser-only. Not a router. - [API — static JSON](https://undominated.ai/api/?format=md): Catalogue, frontier, and LMArena dominance verdicts. No keys. Not a router. - [Shopping-worth spreads — Markdown](https://undominated.ai/spreads/?format=md): Models ranked by actionable multi-seller spread at equal-or-better precision. - [OpenRouter or direct — Markdown](https://undominated.ai/openrouter-vs-direct/?format=md): When routing can find a second price, and when it cannot. Computed counts plus an illustrative break-even; no invented direct list price. - [New releases — Markdown](https://undominated.ai/new/?format=md): Models by the date OpenRouter first listed them, last 90 days, newest first, with rank, price and verdict; unrated stays unrated. - [Model finder — Markdown](https://undominated.ai/finder/?format=md): Three questions (task, budget, must-haves) answered with the leaderboard’s own undominated set; unrated models listed apart; runs client-side. - [Vision input — Markdown](https://undominated.ai/vision/?format=md): Models that accept image input, ranked on the general canon; the frontier is within that set, and unrated models are listed apart. - [Cost calculator — Markdown](https://undominated.ai/calculator/?format=md): Monthly bill per model from your token mix, prompt size and cached share, using the leaderboard’s own tier-ladder pricing. - [Off-peak windows — Markdown](https://undominated.ai/off-peak/?format=md): Models priced by time of day: standard rate, discount windows in UTC and the saving; headlines elsewhere stay standard. - [Price changes — Markdown](https://undominated.ai/price-changes/?format=md): Dated price moves on the balanced blend from the wire, with a weekly tally of moves cheaper and dearer. - [Workflows — Markdown](https://undominated.ai/workflows/?format=md): Reviewed recipes, each pairing a skill, an agent role and an MCP server, with each source and, where established, its licence. - [Licences — Markdown](https://undominated.ai/licences/?format=md): Each licence type with its models, a link to its text and, where the text says so, whether commercial use is allowed. - [Migration guide — Markdown](https://undominated.ai/migrate/?format=md): Every model with a retirement date and its like-for-like replacement: no lost recorded capability, a score at least as high where the model is rated, cheapest at short prompts. - [Open vs closed — Markdown](https://undominated.ai/open-vs-closed/?format=md): Rated open-weight against closed models, with no-status models counted apart, never as closed: best-to-best gap with its error, frontier share, and price at matched score bands. - [Frontier history — Markdown](https://undominated.ai/frontier/history/?format=md): Frontier joins and leaves by day, time-of-day prices marked, frontier rows per published day file beside the board’s own count, and the price-at-floor series. A day file is what the pipeline recorded that day, kept as published: not a replay, and not always the board’s current reading. - [Glossary — Markdown](https://undominated.ai/glossary/?format=md): The terms the leaderboard and the methodology lean on, each defined once and linked to where it is computed. - [The wire — Markdown](https://undominated.ai/wire/?format=md): Material catalogue events only: price moves, context cuts, dated corrections. Not a blog. - [Blind Rank Test — Markdown](https://undominated.ai/blind-rank/?format=md): Adjacent LMArena pairs, guessed then revealed against published error. - [Cheaper alternatives — Markdown](https://undominated.ai/alternatives/?format=md): Dominated models and their alternatives; one page each naming the comparison and its trade. - [Deprecations — Markdown](https://undominated.ai/deprecations/?format=md): Dated retirements plus whether a named successor is a capability superset. - [Traps — Markdown](https://undominated.ai/traps/?format=md): Named pricing failure modes. Tokenizer-tax refuses a percent. - [Disagreement — Markdown](https://undominated.ai/disagreement/?format=md): Where independent boards part company. Arena-only for redistributable claims. - [Context cliffs — Markdown](https://undominated.ai/cliffs/?format=md): Context-tier step functions; every rung kept. - [Significance bands — Markdown](https://undominated.ai/significance/?format=md): Adjacent ranks inside published error; live share. - [Knee — Markdown](https://undominated.ai/knee/?format=md): Where paying more stops buying quality on the frontier. - [Effort pricing — Markdown](https://undominated.ai/effort/?format=md): Mandatory reasoning as a pricing dimension; no τ point estimate. - [Reasoning-token prices — Markdown](https://undominated.ai/reasoning-price/?format=md): Reasoning-token rates as vendors list them, with the context-tier ladder. A missing rate is omitted, not zero. Not a tax estimate. - [Verdict badges — Markdown](https://undominated.ai/badges/?format=md): README badge and expiring frontier hallmark. Alt text is the model name. - [Corrections — Markdown](https://undominated.ai/corrections/?format=md): Published corrections and snapshot changes. - [Cheapest at a quality floor — Markdown](https://undominated.ai/cheapest-at/?format=md): Lowest-priced rated model above a stated capability floor. - [Best models — Markdown](https://undominated.ai/best/?format=md): Capability lenses and significance bands from the same catalogue. - [Providers — Markdown](https://undominated.ai/providers/?format=md): Seller coverage and actionable cross-provider spreads. - [All models — Markdown](https://undominated.ai/models/?format=md): Complete vendor-grouped index linking every publishable model page. ## Machine-readable resources - [Full URL list for agents](https://undominated.ai/llms-full.txt): Complete localized hub and model-page crawl list. - [Catalogue JSON](https://undominated.ai/data/catalogue.json): Models, prices, capability scores, and marketplace signal. - [Plans JSON](https://undominated.ai/data/plans.json): Subscription-plan data. - [Sitemap index](https://undominated.ai/sitemap.xml): Canonical index of localized sitemaps. - [Robots policy](https://undominated.ai/robots.txt): Crawler permissions for search and AI use. ## Optional - [AI crawler guidance](https://undominated.ai/ai.txt): Plain-language guidance complementary to robots.txt.