Question 1

How is the token count calculated?

Accepted Answer

For OpenAI models we run the official BPE tokenizer (`cl100k_base` for GPT-3.5/4, `o200k_base` for GPT-4o) via js-tiktoken — same exact algorithm OpenAI uses. For Claude we use Anthropic's published Claude 2 tokenizer, which matches Claude 3+ within ~5% for English text. For Gemini we approximate at ~4 chars/token because no public Gemini tokenizer exists.

Question 2

Does my text leave the browser?

Accepted Answer

No. All tokenization is client-side. Nothing is sent to any server — not ours, not OpenAI's, not Anthropic's. Open your network tab and verify if you like.

Question 3

Why are input and output prices different?

Accepted Answer

Providers charge more for tokens the model _generates_ than for tokens you _supply_. Output tokens are computationally more expensive because the model decodes them autoregressively. The toggle above lets you project either side of the cost.

Question 4

How much does 1M tokens cost in May 2026?

Accepted Answer

Roughly: **Claude 3.5 Sonnet** is $3 in / $15 out per 1M tokens. **GPT-4o** is $2.50 / $10. **Claude Haiku** is $0.25 / $1.25. **GPT-4o-mini** is $0.15 / $0.60. **Gemini 1.5 Flash** is $0.075 / $0.30. The cheap models are 20-100x cheaper than the flagship models — a big lever for high-volume apps.

Question 5

How accurate are these cost projections?

Accepted Answer

Exact for a known workload — the math is straightforward: (input × input rate) + (output × output rate), with the cached fraction billed at the model's cached-input rate. Pricing comes from each provider's published rate card. The variable is how well your typed numbers match your actual workload. If you don't know your token sizes, the [Token Counter](https://handyai.tools/tools/token-counter/) helps you calibrate against a real sample.

Question 6

What is prompt caching and why does it matter?

Accepted Answer

Anthropic and OpenAI both let you cache the static prefix of a request — typically the system prompt and any retrieved context — and pay a much lower rate on subsequent calls that reuse the same prefix. Anthropic cached reads cost 0.1× the input rate (Sonnet input is $3/M → cached is $0.30/M). OpenAI cached prefixes on GPT-4o-class models cost 0.5×. For a chatbot with a 400-token cached system prompt and 100-token user messages, that's an 80% effective hit rate and a real 5–8× cost reduction.

Question 7

Why doesn't Gemini show a cache discount?

Accepted Answer

Google hasn't published a separate cached-input price for Gemini 1.5 in the rate card we use. The calculator falls back to the full input rate for Gemini, so increasing the cache hit rate doesn't change Gemini's row. If Google adds cached-input pricing, we'll update the table.

Question 8

Are output tokens really billed higher than input?

Accepted Answer

Yes — typically 4–5× higher. Claude 3.5 Sonnet is $3/M input vs $15/M output (5×); GPT-4o is $2.50/M vs $10/M (4×). Output is more expensive because it requires autoregressive decoding (one token at a time). The practical consequence: for workloads where the model writes a lot (summarization, generation), output dominates total cost — which is exactly what the "Summarization" template shows.

Question 9

What's the catch?

Accepted Answer

No sign-up, no email gate, no data leaves your browser unless a tool is explicitly AI-powered (in which case it's clearly marked). The site is funded by ads and a few relevant affiliate links — that's it.

Question 10

Suggesting a tool

Accepted Answer

The tool roadmap is shaped by what people actually search for. If you need a tool that doesn't exist yet, post about it somewhere public and the team will hear about it. (We're keeping it intentionally low-touch — no contact form here.)

HandyAI

About HandyAI

Quick facts about HandyAI

Key Features

Token Management

Cost Estimation

Prompt Analysis

Data Validation

Conversion Tools

Use Cases

Pricing

Integrations(16)

Built for

Frequently asked questions

Reviews (0)

Quick Info

Traffic Sources

Alternatives to HandyAI

Abyss Hub

ACE Studio

Actionbook

Action Sync

Adaapt.AI

Compare HandyAI head-to-head