Know which model to ship.
RankLLMs is an independent leaderboard for 80+ AI models — reasoning, coding, agentic ability, tokens per second, and the API price you’ll actually pay. Interactive charts, head-to-head comparisons, and a public JSON API. Free, no account.
What gets measured.
| Dimension | What it tells you |
|---|---|
| Reasoning | Mathematical and multi-step reasoning scores across benchmark suites. |
| Coding | Code generation and correction accuracy on real programming tasks. |
| Agentic | How well a model sustains multi-step autonomous tool use. |
| Speed | Tokens per second — the number your users feel as latency. |
| API cost | Blended price per million tokens across input and output. |
| Context | The window each model actually accepts, not the marketing figure. |
A leaderboard you can build on.
80+ models tracked
Proprietary and open-weights, scored on the same dimensions — no pay-to-rank listings.
Interactive charts
Intelligence vs. speed, performance vs. price — plot the frontier instead of reading a table.
Head-to-head engine
Any two models, side by side, across every tracked dimension.
Public JSON API
The leaderboard data is served programmatically for your own tools and dashboards.
Benchmark guides
What each benchmark measures, where it misleads, and how to read the scores honestly.
Release coverage
A newsletter and blog tracking model releases and API price changes as they land.
Who it’s for.
AI engineers choosing a production API
The trade you’re making is intelligence against latency against cost. The charts put all three on one axis pair so the decision is arithmetic.
Researchers tracking the frontier
Reasoning and agentic scores across both proprietary and open-weights models, so you can see where the gaps actually are — not where the launch posts say they are.
Developers building on coding agents
Coding accuracy and speed per dollar, filtered to the models that agentic CLI tools actually call.
Straight answers.
- What is RankLLMs?
- An independent leaderboard and comparison platform for large language models. It ranks proprietary and open-weights models on the numbers that decide which one you ship: reasoning, coding accuracy, speed, and real API cost.
- Is RankLLMs free?
- Yes — the leaderboard, the comparison engine, the charts, and the JSON API are all free. There is no account and no paywall.
- What do the benchmarks measure?
- Coding accuracy, mathematical reasoning, agentic ability, tokens-per-second speed, blended API pricing per million tokens, context window, and Arena Elo — each model scored across the set, not cherry-picked.
- Can I compare two models side by side?
- Yes. The comparison engine puts any two models head-to-head across every tracked dimension, so the choice between a fast cheap model and a slow expensive one becomes arithmetic instead of vibes.
- Can I get the data programmatically?
- Yes — a public JSON API serves the leaderboard data, plus an llms.txt file for AI tools that read the site. Pull it into your own dashboards and scripts.
- Who is it for?
- AI engineers choosing a production API, researchers tracking frontier reasoning, and developers comparing the models behind autonomous coding agents.
Stop picking models by vibes.