Guides
What the numbers say
Explainers, analyses, rankings and comparisons, each computed from the index and updated with it.
- Prompt caching: who offers itendpoints that expose a cached-input price650 offers
- Output vs input pricingmedian output price, as a multiple of input3.89×
- Self-host an LLMopen-weight models matched to your hardware109 models
See all 8 ›
- Biggest context windows, and their pricehow much you can feed each model in one callup to 1311k
- Truly open weights vs. open-but-restrictedwhich models you can actually self-host and ship50% permissive
- Knowledge cutoff dates by modelhow current each model's training data isJun 2026
- The age discount: what last season's model costsmedian price by age bracket207 models
- What a micro-SaaS actually costs to runhosting, database, email, payments and the LLM API15 providers
Figures refresh whenever the index rebuilds. Every number traces back to a provider's published price.