Methodology and sources
The original Chinese-language guide was prepared by Codex on 2026-10-21. This independent reference does not represent any model provider. Each model, price, subscription, and benchmark field retains its source URL and verification date. The full research log is in the project’s SOURCE_NOTES.md.
The guide covers selected models from six providers—OpenAI, Anthropic, Google, DeepSeek, Qwen, and xAI—along with individual and team subscriptions and API token prices. The catalog is a curated snapshot; it does not claim to include every model from each provider.
Figures and comparison rules
Section titled “Figures and comparison rules”- API rates are recorded separately for input, cached input, and output, with the listed currency and per-million-token unit.
- Peak and off-peak windows, context thresholds, deployment regions, and price effective dates are noted separately; different rates are not blended into an average.
- Subscriptions distinguish monthly, annual, and per-seat prices. When a current official price is unavailable, the value remains empty and the interface says “No comparable data.”
- Artificial Analysis Intelligence Index scores retain the source model version and reasoning setting. They are not converted to percentages, and Anran does not create its own score.
- Provider recommendations and promotional rankings are not used as numerical metrics.
Updates and limitations
Section titled “Updates and limitations”Public prices and availability can change with time, region, account eligibility, and taxes. A recorded verification date does not guarantee that an account can purchase a plan or that it is available in every country. Check the provider’s official page before purchasing or integrating a model.
This site does not scrape pages automatically, call model APIs, or store API keys. Maintainers update the data manually after checking official material in src/data/catalog.ts and SOURCE_NOTES.md.
Sources
Section titled “Sources”- Official model, price, and subscription pages are listed in the project’s
SOURCE_NOTES.md, which retains full URLs and verification notes. - Public benchmarks: Artificial Analysis LLM Leaderboard and methodology.
- Site template: Astro Starlight Basics, licensed under MIT. The exact template revision is recorded in
SOURCE_NOTES.md.