“Free tier” can mean Gemini works all day without reaching a limit, while Claude hits its limit after a few rounds. The same monthly price can also buy very different capabilities. This guide compares Claude, ChatGPT, Gemini, Grok, and Perplexity from free access through API pricing, then recommends options by use case. Prices follow official pages: Google lists Taiwan prices in TWD, while other providers mainly list USD; check checkout for the amount charged in TWD.
Monthly prices (TWD and USD)
As of 2026-09-29, based on each provider’s official pricing page. Taxes depend on the checkout region.
| Service | Free | Entry paid plan | Main paid plan | High usage |
|---|---|---|---|---|
| Gemini (Google AI) | NT$0 | AI Plus NT$165/month | AI Pro NT$650/month | AI Ultra NT$3,300 or NT$6,500/month |
| ChatGPT | US$0 | Go (US$8/month in the US; local pricing in some regions) | Plus US$20/month | Pro US$100 or US$200/month |
| Claude | US$0 | None | Pro US$20/month (US$17/month billed annually) | Max US$100 or US$200/month |
| Grok | Depends on X account plan | SuperGrok Lite | SuperGrok | SuperGrok Plus, SuperGrok Heavy |
| Perplexity | US$0 | None | Pro (see checkout) | None |
- Google lists Taiwan monthly prices in TWD (NT$165/650/3,300/6,500) on its official plan page.
- New subscriptions and upgrades to ChatGPT Pro US$200 have been paused since 2026-09-10. Existing subscribers continue to renew, and Pro US$100 is unaffected (OpenAI details).
- Claude lists local-currency billing; check checkout for the amount in TWD.
Free-tier comparison
| Model | What you can use for free | Quota | Biggest limitation |
|---|---|---|---|
| Claude | Sonnet + Haiku | Medium; wait for the quota to reset after reaching the message limit | Flagship models and Claude Code require payment |
| ChatGPT | Primary model, unlimited text | Generous | Images and advanced tools have limits |
| Gemini | Flash + image generation + Deep Research + Gemini Live | Very generous | Flagship models and video generation require payment |
| Grok | Depends on X account plan | Variable | Without an X account, there is effectively no free tier |
| Perplexity | Unlimited basic search + limited Pro Search | Sufficient | Deep Research requires Pro |
Gemini offers the widest range of free features: the primary Flash model, Nano Banana 2 image generation, Gemini Live voice, and Deep Research. ChatGPT offers unlimited text conversations, but limits images and tools. Claude has the highest quality but the tightest quota. Perplexity is the best starting point for research because its basic search is strong.
A zero-cost stack: ChatGPT for everyday text, Gemini for visuals, Perplexity for search, and Claude’s limited quota for high-quality writing. That covers about 80% of everyday needs.
Paid-tier comparison
| Category | Claude Pro/Max | ChatGPT Go/Plus/Pro | Google AI Plus/Pro/Ultra | Grok / SuperGrok | Perplexity Pro |
|---|---|---|---|---|---|
| Monthly price | US$20 / Max US$100+ | Go US$8; Plus US$20; Pro US$100–200 | NT$165 / NT$650 / NT$3,300–6,500 | See checkout (Lite / SuperGrok / Plus / Heavy) | See checkout page |
| Flagship model | Opus; Fable costs extra usage on Pro and uses up to half the weekly quota on Max | GPT-6 Pro (Pro only); Plus includes GPT-6 Astra in Work and Codex | 3.1 Pro, 3.8 Flash (AI Pro and above) | Grok 4.6 | Search integration |
| Images | None | Yes | Yes (Nano Banana 2) | Yes | None |
| Extras | Claude Code, Research | GPTs, Codex, Deep Research | Workspace integration, video generation, Flow credits, Antigravity | X integration | File analysis, Deep Research |
Claude: flagship models justify the subscription
Opus breaks down assumptions and analyzes problems from multiple angles where Sonnet may simply provide a reasonable answer. Fable is the new flagship for reasoning and writing. Pro costs US$20/month, or US$17/month with annual billing, and includes Opus. Fable is not included in Pro’s plan quota and requires paid usage credits. Max starts at US$100 for higher limits; Fable can use up to half of the weekly quota on Max (Claude pricing). See Claude Opus vs Sonnet for details.
ChatGPT: everything in one plan
Go costs US$8/month in the US (OpenAI says it may test ads in the future) and suits occasional users who reach the free limit. Its model is the same GPT-5.6 Luna as Free, with a higher quota. Plus costs US$20/month and includes search, images, voice, data analysis, and GPTs, with GPT-5.6 Sol for chat. GPT-6 Pro, powered by GPT-6 Astra, is available in Chat only on Pro; Plus users can use GPT-6 Astra in ChatGPT Work and Codex (OpenAI details). ChatGPT Plus is not the leader in every individual category, but it is the only option that does nearly everything. See Claude vs ChatGPT and Gemini vs ChatGPT.
Gemini: the cheapest paid entry
AI Plus costs NT$165/month, the lowest entry price of the five, and includes video generation, Flow credits, and 400 GB of storage. AI Pro costs NT$650/month and adds 3.1 Pro, 3.8 Flash, the Antigravity coding tool, and 5 TB. For people who already need Google One, AI Plus can feel like an almost-free AI upgrade. Its disadvantage is a less complete Chinese experience than ChatGPT. See Gemini Free vs Google AI Pro.
Perplexity Pro: built for verification
Perplexity does not compete on chat quality. Its value is search depth and source credibility. Pro substantially increases query limits and adds file analysis and deeper multi-source research. See Perplexity vs ChatGPT.
API pricing
APIs charge by token, and prices change frequently. Check the latest platform pages for exact figures. Output tokens typically cost three to five times as much as input tokens.
| Platform | Representative model | Input / output per 1M tokens | Notes |
|---|---|---|---|
| Claude | Fable 5.1 | US$10 / US$50 | Opus 5.5 US$4 / US$20, Sonnet 5.5 US$2 / US$10, Haiku 4.5 US$1 / US$5; prompt caching available |
| OpenAI | GPT-6 Sol | US$2 / US$10 | GPT-6 Luna US$0.10 / US$0.50; GPT-6 Astra US$10 / US$50; batch API available |
| Gemini | 3.8 Flash | US$0.75 / US$3.75 (introductory price) | Generous free quota; small personal projects may cost nothing |
| Grok | Grok 4.7 | US$2 / US$6 | A fast version is also available at twice the speed and price |
The first rule: set a hard limit on every API key. A buggy pipeline can repeat the same call hundreds of times and create a large bill. Use batch processing to lower costs when the workflow does not require an immediate response.
Best value by use case
- Students and light users: ChatGPT + Gemini + Perplexity free tiers, at zero cost. See the Gemini student plan guide for student offers.
- Content creators: Claude Pro for text + ChatGPT Plus for images. If the budget covers only one, choose ChatGPT and use Claude’s free quota for writing.
- Developers: Claude Max for architecture and code review + ChatGPT Plus or Pro for Codex execution.
- Heavy users: Claude Max + ChatGPT Pro + Google AI Pro + Perplexity Pro + Grok. When the tools save billable time, the return is often positive.
If only one can remain, choose Claude Max. Its writing and coding advantages are the hardest to replace; Gemini Free can cover images and Perplexity Free can cover search.
Penchan’s experience
I subscribe to all five: Claude Max, ChatGPT Pro, Google AI Pro, Perplexity Pro, and Grok. Claude handles long-form and multi-agent work, ChatGPT handles ideation and programming, Gemini creates visuals, Perplexity verifies facts, and Grok tracks real-time information. If I could keep only one, it would be Claude Max because the largest gaps are in writing and programming, while the other functions have free substitutes.
Google’s TWD prices are official list prices; other providers’ USD prices are official list prices, and TWD charges depend on regional checkout. Actual prices follow each platform’s latest announcements. This consumer comparison of AI subscriptions is not investment advice.