Question 1

What is a token in large language models?

Accepted Answer

A token is a small chunk of text that a model reads and generates as its basic unit. Your prompt and the model’s reply are both broken into tokens before processing. The exact split depends on the model’s tokenizer, so the same words can produce slightly different token counts on different systems.

Question 2

What is tokenization?

Accepted Answer

Tokenization is the step that turns raw text into a sequence of tokens the model can work with. It is handled by a tokenizer tied to a specific model family. That is why “tokens” are not the same as words or characters, even though they are related.

Question 3

Why does token count matter for API costs?

Accepted Answer

Most hosted LLM APIs bill by how much of the model’s capacity you use, and that usage is tracked in tokens. Longer prompts, larger system instructions, tool definitions, and longer replies all increase what you send through the API, which affects your bill.

Question 4

What is a context window?

Accepted Answer

The context window is how much text the model can consider at once for a single request—your input plus anything the model has already generated in that turn. If you exceed it, you may need to trim content, summarize, or split work across multiple calls.

Question 5

How do input and output pricing differ?

Accepted Answer

Providers usually list separate rates for tokens you send in (input) and tokens the model produces (output). Output is often priced higher because generation is more compute-intensive. Your total charge reflects both sides of the conversation.

Question 6

What is cached input pricing?

Accepted Answer

Some providers can reuse recently seen input segments instead of reprocessing them from scratch. When that applies, those tokens may be billed at a reduced cached input rate rather than the standard input rate. Whether caching helps depends on the provider, the API, and how your requests are structured.

Question 7

How can I reduce LLM API costs in general?

Accepted Answer

Keep prompts focused, avoid repeating large blocks of text on every call when you can reuse or reference them, choose a model that matches the task difficulty, and cap maximum reply length when appropriate. Reviewing usage in your provider dashboard also helps spot spikes early.

Question 8

Why do different models have different costs?

Accepted Answer

Pricing reflects model capability, size, training and serving expense, and commercial positioning. Stronger or more specialized models typically cost more per token than smaller or general-purpose ones. Providers adjust rates over time as offerings evolve.

Question 9

Is an online token count the same as my provider’s bill?

Accepted Answer

A browser-based counter gives a useful estimate, especially for plain text. Your invoice may still differ because providers can count tokens with their own tokenizer version, and real requests can include hidden formatting, special tokens, or attachments that add tokens beyond what you pasted here.

Question 10

Does this tool send my text to a server?

Accepted Answer

No. Token counting and cost estimates run entirely in your browser. Your text stays on your device unless you choose to share it elsewhere.

Question 11

What is “Caveman” on the Prompt Optimizer tab?

Accepted Answer

It is a fourth compression preset that runs Gentle, Balanced, and Aggressive steps, then applies an extra terse pass (heuristic trims like filler adverbs). It is inspired by the idea of ultra-compact prose and is not the same as the standalone Caveman CLI or plugin suite from the open-source Caveman project—those tools use different pipelines and run outside this page.

Question 12

How does Prompt Optimizer relate to the Caveman toolkit (CLI / editor modes)?

Accepted Answer

The Caveman project (for example caveman-compress and editor modes) targets assistant responses or project memory files with its own rules. This site’s Prompt Optimizer is a self-contained, browser-only heuristic—you can use both ideas together conceptually, but they do not share code or identical output.

Model	Provider	Context window	Input cost	Cached input cost	Output cost
GPT-5.4	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5.2	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1 mini	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1 nano	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5 nano	OpenAI	1M	$0.00	—	$0.00
o3	OpenAI	200K	$0.00	$0.00	$0.00
o3-pro	OpenAI	200K	$0.00	—	$0.00
o4-mini	OpenAI	200K	$0.00	$0.00	$0.00
Claude Opus 4.6	Anthropic	200K	$0.00	$0.00	$0.00
Claude Opus 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4.6	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4	Anthropic	200K	$0.00	$0.00	$0.00
Claude Haiku 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Haiku 3.5	Anthropic	200K	$0.00	$0.00	$0.00
Gemini 3.1 Pro	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Pro	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Flash	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Flash-Lite	Google	1M	$0.00	—	$0.00
Gemini 2.0 Flash	Google	1M	$0.00	$0.00	$0.00
Grok 3	xAI	128K	$0.00	—	$0.00
Grok 3 mini	xAI	128K	$0.00	—	$0.00
DeepSeek V3	DeepSeek	128K	$0.00	$0.00	$0.00
DeepSeek R1	DeepSeek	128K	$0.00	$0.00	$0.00
Llama 4 Maverick	Meta	1M	$0.00	—	$0.00
Llama 4 Scout	Meta	512K	$0.00	—	$0.00
Llama 3.3 70B	Meta	128K	$0.00	—	$0.00
Mistral Large	Mistral	128K	$0.00	—	$0.00
Mistral Small	Mistral	128K	$0.00	—	$0.00
Codestral	Mistral	256K	$0.00	—	$0.00

Model	Provider	Context window	Input cost	Cached input cost	Output cost
GPT-5.4	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5.2	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1 mini	OpenAI	1M	$0.00	$0.00	$0.00
GPT-4.1 nano	OpenAI	1M	$0.00	$0.00	$0.00
GPT-5 nano	OpenAI	1M	$0.00	—	$0.00
o3	OpenAI	200K	$0.00	$0.00	$0.00
o3-pro	OpenAI	200K	$0.00	—	$0.00
o4-mini	OpenAI	200K	$0.00	$0.00	$0.00
Claude Opus 4.6	Anthropic	200K	$0.00	$0.00	$0.00
Claude Opus 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4.6	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Sonnet 4	Anthropic	200K	$0.00	$0.00	$0.00
Claude Haiku 4.5	Anthropic	200K	$0.00	$0.00	$0.00
Claude Haiku 3.5	Anthropic	200K	$0.00	$0.00	$0.00
Gemini 3.1 Pro	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Pro	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Flash	Google	1M	$0.00	$0.00	$0.00
Gemini 2.5 Flash-Lite	Google	1M	$0.00	—	$0.00
Gemini 2.0 Flash	Google	1M	$0.00	$0.00	$0.00
Grok 3	xAI	128K	$0.00	—	$0.00
Grok 3 mini	xAI	128K	$0.00	—	$0.00
DeepSeek V3	DeepSeek	128K	$0.00	$0.00	$0.00
DeepSeek R1	DeepSeek	128K	$0.00	$0.00	$0.00
Llama 4 Maverick	Meta	1M	$0.00	—	$0.00
Llama 4 Scout	Meta	512K	$0.00	—	$0.00
Llama 3.3 70B	Meta	128K	$0.00	—	$0.00
Mistral Large	Mistral	128K	$0.00	—	$0.00
Mistral Small	Mistral	128K	$0.00	—	$0.00
Codestral	Mistral	256K	$0.00	—	$0.00

LLM Token Calculator

Related Tools

LLM RAM Calculator

Token Speed Simulator

Frequently Asked Questions About LLM Tokens

LLM Token Calculator

Related Tools

LLM RAM Calculator

Token Speed Simulator

Frequently Asked Questions About LLM Tokens