Editorial policy
How we check facts, rank models, label sponsorships, use AI tools and correct mistakes.
Our standards
We publish information people use to spend money and make technical decisions, so accuracy comes first. We prefer primary sources: provider documentation and pricing pages, model cards, official announcements and research papers. We say what we don't know, date anything that changes, and keep opinion clearly separate from fact.
How we verify model data
- Prices, context windows, output limits, licenses and release dates come from the provider's official pages. Where possible we cross-check with a second source.
- Prices are list rates for the standard tier in US dollars per million tokens. Long-prompt surcharges, discounts and regional pricing are noted on the model page.
- Every model page lists the sources we used and the date we checked them.
- If we can't confirm a value, we leave it blank and explain why. We don't estimate published specs.
- We re-check data regularly and whenever a provider announces a change.
How tiers and rankings work
Each model has a capability tier: frontier, advanced, efficient or compact. Tiers are an editorial grouping based on how the provider positions the model and where it stands on public benchmarks and arena rankings at the time of review. They are not results of our own testing, and we label them that way wherever they appear.
Our rankings page uses a published formula that combines the tier with price, context window, openness, input types and release date. You can change the weights yourself. Shortlists order models by status, tier and price unless the page says otherwise. No one can pay to change a tier, a score or an order.
How we use AI
We use AI tools, including large language models, to help research, draft, edit and write code for this site. People decide what we publish. Facts on every page are checked against the sources listed on that page, and model data is never taken from an AI model's memory. When a guide shows example model output, it's presented as an example.
Sponsored content and advertising
- Display ads are served by Google AdSense. We don't choose individual ads.
- Direct sponsorships and featured listings are labeled as sponsored and appear only in fixed sponsor slots.
- Sponsored articles are labeled at the top, follow these standards and may not make claims we can't check.
- Sponsored and affiliate links carry rel="sponsored".
- Sponsors never see or approve editorial content before it's published, and never affect rankings or verdicts.
How we make money
LLModel is funded by display advertising, labeled sponsorships, reader support and referral fees from implementation partners. Referral fees apply only when a reader asks to be introduced to a partner and later hires them; partners are chosen for fit with the project, not for the fee. If we add affiliate links to the resources page, they're marked and use rel="sponsored". Editorial picks there are never paid for.
Independence
We aren't owned by, funded by or affiliated with any model provider. If a provider ever gives us free credits for testing, we'll say so on the pages that use them. We don't accept payment in exchange for coverage.
Curated videos and external links
We embed YouTube videos we think are useful and summarize them in our own words. We aren't affiliated with the creators, and embedding a video isn't an endorsement of everything in it. External links point to the most authoritative source we can find.
Corrections
If you find an error, please tell us with the page address and, ideally, a source. We fix confirmed errors promptly and update the page's checked date. Significant corrections are noted on the page.