mofcloud
English
Global site
English mofcloud.com Current site
China site
中文 mofcloud.cn

LLM Token & Cost Calculator

1Paste your prompt or text

0
CJK chars
0
EN words
0
Other
0
≈ input tokens

Rough: 1 CJK char ≈ 1.3 tokens, 1 English word ≈ 1.3 tokens. Real counts vary by tokenizer (±10%).

2Cost parameters

Input tokens (sent per request)116
Output tokens (generated per request)800
Thinking tokens (reasoning models · billed at output rate)0
Calls per day100 / day → 3,000 / mo
Cache hit rate (repeated prefix billed at cache price)0%

Cost across models

# Input / Output / 1M

See the full model price comparison →

Guide

How is token cost calculated?

Costing an LLM API call is really just two steps. Per-call = input tokens × input price ＋ output tokens × output price (prices are quoted per million tokens); then × calls per day × 30 for a monthly budget. This tool does both for you, and lines every model up side by side.

Why enter input and output separately? Because almost every vendor prices output higher than input, often 2–4× — generating costs more compute than reading. So the same 10k tokens cost very differently for “long input, short output” (summarizing) versus “short input, long output” (writing). A single blended price hides that.

A worked example

A support bot: ~800 input and ~200 output tokens per call, 30,000 calls a day. Say a model charges $1 input / $4 output per million tokens —
per call = (800×1 ＋ 200×4) ÷ 1M = $0.0016; a month ≈ 0.0016 × 30k × 30 = $1,440.
Switch to a model with cheaper output and that monthly bill often halves — exactly what the table above is for.

Chinese models (Doubao / Qwen / ERNIE / Kimi, etc.) use each vendor’s official CNY price; global models convert at the live rate, toggled ￥ / $. For each model’s context window, capability tags and more, see the LLM price comparison.

FAQ

What's the difference between input and output tokens?

Input tokens are what you send the model (system prompt, user question, retrieved context); output tokens are consumed as the model generates its reply. Output is usually priced higher than input.

How many characters is one token?

Roughly: 1 English word ≈ 1.3 tokens, 1 CJK character ≈ 1.3 tokens. Tokenizers vary by vendor; official billing is authoritative.

Why show a monthly cost from a daily call count?

Budgets are usually monthly. We multiply per-call cost by calls/day and by 30 days to turn a unit price into a team- or product-level monthly budget.

What is the cache hit rate and why does it save money?

Many models offer a discounted 'context cache' price for repeated input prefixes (a fixed system prompt, knowledge-base context). Raise the hit rate and that share of input tokens bills at the cache price. Only models that publish a cache price are affected; the rest bill at the normal input price.

How are thinking tokens billed?

Reasoning models emit thinking tokens that are normally billed at the same rate as output tokens. We list them separately so you can estimate the extra cost of reasoning models; they are folded into output cost in the total.

Are Chinese models priced in CNY or USD?

Chinese models (Doubao / ERNIE / Hunyuan / Spark / Qwen, etc.) use each vendor's official CNY price; global models use USD. Toggle ￥ / $ to convert at the latest rate.

Which models are covered? Where can I see more?

The calculator reuses the main pricing dataset covering major Chinese & global models. For the full list with context windows and capabilities, see the LLM price comparison.