LLM Abacus is a price table for large language model APIs. It covers 57 models from 16 vendors, including OpenAI, Anthropic, Google, xAI, DeepSeek, Zhipu, Moonshot, Alibaba, ByteDance, Baidu, Tencent and MiniMax, with input, output and cached-input prices side by side. Overseas vendors keep their USD list price and are converted through a single exchange-rate constant, so both currencies stay visible. Models priced by context length show every tier, and models with peak and off-peak pricing show both bands. An automated job re-checks prices every day against the vendors' own pages and flags anything that has drifted; the changelog has RSS and JSON feeds.

Calculators: paste a prompt and get the token count and cost per model; a monthly bill calculator; a routing calculator that mixes a cheap model with a flagship one and shows what the split saves; a subscription versus API break-even for heavy Claude Code use; a local hardware versus cloud API break-even; an agent cost simulator for multi-turn context growth with prompt caching; and a use-case recommender that ranks models by cost and fit for a task. There is also an embedding model price table, a channel comparison of official prices against OpenRouter, Volcengine and Alibaba Bailian, a relay API checker that measures latency and the model a relay actually reports (your key never leaves the browser), a ratio converter for relay pricing, a price history timeline, and a live latency board with first-token latency and output speed, exportable as CSV.

For programs there is a JSON price API and an MCP server. The site is Chinese-first with English pages, plus about two dozen explanatory articles.