Appearance

Language

AI subscription vs API cost (NT$)

Every AI subscription side by side in NT$, API bills from your token usage, Chinese characters to tokens, and the usage where a plan and a model cost the same. Prices checked 2026-11-16 on the official pricing pages.

Default: Bank of Taiwan USD spot selling rate 32.91 (posted 2026-11-15 09:25). Card charges use the card network's rate. Tax or your card’s foreign-transaction fee, whatever your bill adds to the US$ price. Not applied to plans with an official NT$ price.
Sort

Subscriptions in NT$ a month

PlanUS$ a monthNT$ a monthNT$ fromNote
ChatGPT
GoUS$8.00NT$263exchange rate
PlusUS$20NT$658exchange rate
Pro $100US$100NT$3,291exchange rate
Pro $200US$200NT$6,582exchange rate
Pro $500US$500NT$16,455exchange rate
Claude
ProUS$20NT$658exchange rateUS$17/mo billed annually (US$200 up front)
MaxUS$100NT$3,291exchange rate
UltraUS$200NT$6,582exchange rate
Cursor
PlusUS$8.00NT$263exchange rate
PremiumUS$30NT$987exchange rate
SuperUS$100NT$3,291exchange rate
UltraUS$200NT$6,582exchange rate
Devin
ProUS$20NT$658exchange rate
MaxUS$200NT$6,582exchange rate
GitHub Copilot
ProUS$10NT$329exchange rate
Pro+US$39NT$1,283exchange rate
MaxUS$100NT$3,291exchange rate
Google AI
PlusUS$4.99NT$165official NT$ price
ProUS$20NT$650official NT$ price
UltraUS$100NT$3,300official NT$ price
Ultra 20xUS$200NT$6,500official NT$ price

API, paid by usage

Usage per
Long prompts you send again (system prompt, code, documents) are billed at the lower cached price when they hit the cache. Not sure? Enter 0.

Estimate: OpenAI’s public tokenizers measured on 53 of this site’s Traditional Chinese posts give 0.66–1.25 tokens per character. Claude, Gemini and others use different tokenizers; use each vendor’s token-counting API for an exact count.

2M input and 500K output tokens a month US$ per million tokens

ModelInputOutputCached inputYour usage a monthNote
Anthropic
Claude Fable 5.2$8$40$0.2NT$1,185US$36
Claude Haiku 5.5$0.1$0.5$0.01NT$15US$0.45Prompts up to 100K tokens
Claude Opus 5.5$4$20$0.2NT$592US$18
Claude Sonnet 5.5$2$10$0.1NT$296US$9.00
DeepSeek
DeepSeek Flash$0.3$1.2$0.006NT$39US$1.20Peak-hour price; off-peak is half
DeepSeek V4 Pro$1.32$3.96$0.044NT$152US$4.62Peak-hour price; off-peak is half
Google
Gemini 3.5 Flash-Lite$0.3$2.5$0.03NT$61US$1.85
Gemini 3.8 Flash$0.75$3.75$0.075NT$111US$3.38From 2027-01-01: 1.5 in / 7.5 out; cache storage billed separately
Gemini 4 Carbon$2$10$0.1NT$296US$9.00From 2026-12-01: 4 in / 20 out; cache storage billed separately
Gemini 4 Hydrogen$1.5$7.5$0.075NT$222US$6.75
OpenAI
GPT-6 Astra$10$50$1NT$1,481US$45
GPT-6 Luna$0.1$0.5$0.01NT$15US$0.45
GPT-6.1 Sol$2$10$0.1NT$296US$9.00
GPT-6.1 Terra$1$5$0.05NT$148US$4.50
xAI
Grok 4.3$1.25$2.5$0.2NT$123US$3.75
Grok 4.7$2$6$0.5NT$230US$7.00
Grok Build 0.1$1$2$0.2NT$99US$3.00

At what usage does the API cost as much as a plan?

Pick a plan and a model to see where they cost the same at your input/output mix.

Prices checked 2026-11-16 on each vendor’s official pricing page, before tax; vendors can change them at any time. Not listed: xAI SuperGrok and Perplexity Pro (no readable price on the official page), and ChatGPT’s NT$ prices (no price in the official page’s text).

Price sources