AI Token Calculator

Tokenizer

Upload file

Drag & drop a text, code or image file here, or use “Upload file”. Text & code (.txt, .md, .csv, .json, .yaml, .xml, .html, .js, .ts, .py, .php, .sql, …) are counted as text; images (.png, .jpg, .webp, .gif) are measured as vision tokens. Max 10 MB.

Private by design. Your text, files and images are processed entirely in your browser. Nothing is sent to our servers, stored, or logged. Reset with “Clear” to wipe everything from memory.

Counts

Tokens: 0
Characters: 0
Characters (no spaces): 0
Words: 0
Lines: 0
Sentences: 0

Token counts use OpenAI's BPE tokenizers where available; otherwise a character-based estimate (shown with “≈”). Image files are measured as high-detail vision tokens. Other providers tokenize differently, so treat non-OpenAI models as approximate.

Token visualization

Pieces appear as the local tokenizer reads your text.

0 tokens

Paste or type text above to see how it is split into tokens.

Search models

Provider

Price Comparison

Estimated input cost for your current 0 tokens, sorted from highest to lowest.

Highest

— —

Lowest

— —

$0Max —

Add some text above to estimate cost across models.

Full model pricing table

Rates update the estimates as soon as you paste text.

Model	Provider	Context	Current rate / 1M tokens	Estimated cost	Pricing tier
babbage-002	OpenAI	16K	Input $0.4 Cached Not offered Output $0.4	Input — Cached — Output —	Standard
chatgpt-4o-latest	OpenAI	128K	Input $5 Cached Not offered Output $15	Input — Cached — Output —	Standard
chatgpt-image-latest	OpenAI	128K	Input $5 Cached $1.25 Output $10	Input — Cached — Output —	Standard
Claude Haiku 3	Anthropic	200K	Input $0.25 Cached $0.03 Output $1.25	Input — Cached — Output —	Standard
Claude Haiku 3.5	Anthropic	200K	Input $0.8 Cached $0.08 Output $4	Input — Cached — Output —	Standard
Claude Haiku 4.5	Anthropic	200K	Input $1 Cached $0.1 Output $5	Input — Cached — Output —	Standard
Claude Haiku 4.5	Perplexity	200K	Input $1 Cached $0.1 Output $5	Input — Cached — Output —	Standard
Claude Opus 3	Anthropic	200K	Input $15 Cached $1.5 Output $75	Input — Cached — Output —	Standard
Claude Opus 4	Anthropic	200K	Input $15 Cached $1.5 Output $75	Input — Cached — Output —	Standard
Claude Opus 4.1	Anthropic	200K	Input $15 Cached $1.5 Output $75	Input — Cached — Output —	Standard
Claude Opus 4.5	Perplexity	200K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Opus 4.5	Anthropic	200K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Opus 4.6	Anthropic	200K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Opus 4.6	Perplexity	200K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Opus 4.7	Anthropic	200K / 32K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Opus 4.7	Perplexity	200K / 32K	Input $5 Cached $0.5 Output $25	Input — Cached — Output —	Standard
Claude Sonnet 3.7	Anthropic	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
Claude Sonnet 4	Anthropic	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
Claude Sonnet 4.5	Anthropic	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
Claude Sonnet 4.5	Perplexity	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
Claude Sonnet 4.6	Anthropic	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
Claude Sonnet 4.6	Perplexity	200K	Input $3 Cached $0.3 Output $15	Input — Cached — Output —	Standard
codex-mini-latest	OpenAI	200K / 100K	Input $1.5 Cached $0.375 Output $6	Input — Cached — Output —	Standard
computer-use-preview	OpenAI	—	Input $3 Cached Not offered Output $12	Input — Cached — Output —	Standard
davinci-002	OpenAI	16K	Input $2 Cached Not offered Output $2	Input — Cached — Output —	Standard
gemini-2.5-pro	Perplexity	1M	Input $1.25 Cached Not offered Output $10	Input — Cached — Output —	Standard Long context after 200K input: input $2.5, output $15.
gemini-3-flash-preview	Perplexity	1M	Input $0.5 Cached Not offered Output $3	Input — Cached — Output —	Standard
gemini-3.1-pro-preview	Perplexity	1M	Input $2 Cached Not offered Output $12	Input — Cached — Output —	Standard Long context after 200K input: input $4, output $18.
gpt-3.5-0301	OpenAI	4K	Input $1.5 Cached Not offered Output $2	Input — Cached — Output —	Standard
gpt-3.5-turbo	OpenAI	16K	Input $0.5 Cached Not offered Output $1.5	Input — Cached — Output —	Standard
gpt-3.5-turbo Legacy	OpenAI	16K	Input $3 Cached Not offered Output $6	Input — Cached — Output —	Standard
gpt-3.5-turbo-0125	OpenAI	16K	Input $0.5 Cached Not offered Output $1.5	Input — Cached — Output —	Standard
gpt-3.5-turbo-0613	OpenAI	4K	Input $1.5 Cached Not offered Output $2	Input — Cached — Output —	Standard
gpt-3.5-turbo-1106	OpenAI	16K	Input $1 Cached Not offered Output $2	Input — Cached — Output —	Standard
gpt-3.5-turbo-16k-0613	OpenAI	16K	Input $3 Cached Not offered Output $4	Input — Cached — Output —	Standard
gpt-3.5-turbo-instruct	OpenAI	4K	Input $1.5 Cached Not offered Output $2	Input — Cached — Output —	Standard
gpt-4-0125-preview	OpenAI	128K	Input $10 Cached Not offered Output $30	Input — Cached — Output —	Standard
gpt-4-0314	OpenAI	8K	Input $30 Cached Not offered Output $60	Input — Cached — Output —	Standard
gpt-4-0613	OpenAI	8K	Input $30 Cached Not offered Output $60	Input — Cached — Output —	Standard
gpt-4-1106-preview	OpenAI	128K	Input $10 Cached Not offered Output $30	Input — Cached — Output —	Standard
gpt-4-1106-vision-preview	OpenAI	128K	Input $10 Cached Not offered Output $30	Input — Cached — Output —	Standard
gpt-4-32k	OpenAI	33K	Input $60 Cached Not offered Output $120	Input — Cached — Output —	Standard
gpt-4-turbo-2024-04-09	OpenAI	128K	Input $10 Cached Not offered Output $30	Input — Cached — Output —	Standard
gpt-4.1	OpenAI	1,050K / 33K	Input $2 Cached $0.5 Output $8	Input — Cached — Output —	Standard
gpt-4.1-2025-04-14	OpenAI	1,000K	Input $3 Cached $0.75 Output $12	Input — Cached — Output —	Standard
gpt-4.1-mini	OpenAI	1,050K / 33K	Input $0.4 Cached $0.1 Output $1.6	Input — Cached — Output —	Standard
gpt-4.1-mini-2025-04-14	OpenAI	1M	Input $0.8 Cached $0.2 Output $3.2	Input — Cached — Output —	Standard
gpt-4.1-nano	OpenAI	1,050K / 33K	Input $0.1 Cached $0.025 Output $0.4	Input — Cached — Output —	Standard
gpt-4.1-nano-2025-04-14	OpenAI	1M	Input $0.2 Cached $0.05 Output $0.8	Input — Cached — Output —	Standard
gpt-4o	OpenAI	128K / 16K	Input $2.5 Cached $1.25 Output $10	Input — Cached — Output —	Standard
gpt-4o-2024-05-13	OpenAI	128K / 4K	Input $5 Cached Not offered Output $15	Input — Cached — Output —	Standard
gpt-4o-2024-08-06	OpenAI	128K	Input $3.75 Cached $1.875 Output $15	Input — Cached — Output —	Standard
gpt-4o-audio-preview	OpenAI	128K / 4K	Input $2.5 Cached Not offered Output $10	Input — Cached — Output —	Standard
gpt-4o-mini	OpenAI	128K / 16K	Input $0.15 Cached $0.075 Output $0.6	Input — Cached — Output —	Standard
gpt-4o-mini-2024-07-18	OpenAI	128K	Input $0.3 Cached $0.15 Output $1.2	Input — Cached — Output —	Standard
gpt-4o-mini-audio-preview	OpenAI	128K / 16K	Input $0.15 Cached Not offered Output $0.6	Input — Cached — Output —	Standard
gpt-4o-mini-realtime-preview	OpenAI	128K / 16K	Input $0.6 Cached $0.3 Output $2.4	Input — Cached — Output —	Standard
gpt-4o-mini-search-preview	OpenAI	128K / 16K	Input $0.15 Cached Not offered Output $0.6	Input — Cached — Output —	Standard
gpt-4o-mini-tts	OpenAI	—	Input $0.6 Cached Not offered Output Not offered	Input — Cached — Output —	Standard
gpt-4o-realtime-preview	OpenAI	128K / 16K	Input $5 Cached $2.5 Output $20	Input — Cached — Output —	Standard
gpt-4o-search-preview	OpenAI	128K / 16K	Input $2.5 Cached Not offered Output $10	Input — Cached — Output —	Standard
gpt-5	OpenAI	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5-chat-latest	OpenAI	128K / 16K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5-codex	OpenAI	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5-mini	Perplexity	400K / 128K	Input $0.25 Cached $0.025 Output $2	Input — Cached — Output —	Standard
gpt-5-mini	OpenAI	400K / 128K	Input $0.25 Cached $0.025 Output $2	Input — Cached — Output —	Standard
gpt-5-nano	OpenAI	400K / 128K	Input $0.05 Cached $0.005 Output $0.4	Input — Cached — Output —	Standard
gpt-5-pro	OpenAI	400K / 128K	Input $15 Cached Not offered Output $120	Input — Cached — Output —	Standard
gpt-5-search-api	OpenAI	128K / 16K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
GPT-5.1	Perplexity	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
GPT-5.1	OpenAI	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5.1-chat-latest	OpenAI	128K / 16K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5.1-codex	OpenAI	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5.1-codex-max	OpenAI	400K / 128K	Input $1.25 Cached $0.125 Output $10	Input — Cached — Output —	Standard
gpt-5.1-codex-mini	OpenAI	400K / 100K	Input $0.25 Cached $0.025 Output $2	Input — Cached — Output —	Standard
gpt-5.2	Perplexity	400K / 128K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.2	OpenAI	400K / 128K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.2-chat-latest	OpenAI	128K / 16K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.2-codex	OpenAI	400K / 128K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.2-pro	OpenAI	400K / 128K	Input $21 Cached Not offered Output $168	Input — Cached — Output —	Standard
gpt-5.3-chat-latest	OpenAI	400K / 128K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.3-codex	OpenAI	400K / 128K	Input $1.75 Cached $0.175 Output $14	Input — Cached — Output —	Standard
gpt-5.4	Perplexity	1M / 128K	Input $2.5 Cached $0.25 Output $15	Input — Cached — Output —	Standard
gpt-5.4	OpenAI	1M / 128K	Input $2.5 Cached $0.25 Output $15	Input — Cached — Output —	Standard Long context after 272K input: input $5, cached $0.5, output $22.5.
gpt-5.4-mini	Perplexity	1M / 128K	Input $0.75 Cached Not offered Output $4.5	Input — Cached — Output —	Standard
gpt-5.4-mini	OpenAI	1M / 128K	Input $0.75 Cached $0.075 Output $4.5	Input — Cached — Output —	Standard
gpt-5.4-nano	OpenAI	1M / 128K	Input $0.2 Cached $0.02 Output $1.25	Input — Cached — Output —	Standard
gpt-5.4-nano	Perplexity	1M / 128K	Input $0.2 Cached Not offered Output $1.25	Input — Cached — Output —	Standard
gpt-5.4-pro	OpenAI	1M / 128K	Input $30 Cached Not offered Output $180	Input — Cached — Output —	Standard Long context after 272K input: input $60, output $270.
gpt-5.5	OpenAI	1,050K / 128K	Input $5 Cached $0.5 Output $30	Input — Cached — Output —	Standard Long context after 272K input: input $10, cached $1, output $45.
gpt-5.5-pro	OpenAI	1,050K / 128K	Input $30 Cached Not offered Output $180	Input — Cached — Output —	Standard Long context after 272K input: input $60, output $270.
gpt-5.6-luna	OpenAI	1,050K / 128K	Input $1 Cached $0.1 Output $6	Input — Cached — Output —	Standard Long context after 272K input: input $2, cached $0.2, output $9.
gpt-5.6-sol	OpenAI	1,050K / 128K	Input $5 Cached $0.5 Output $30	Input — Cached — Output —	Standard Long context after 272K input: input $10, cached $1, output $45.
gpt-5.6-terra	OpenAI	1,050K / 128K	Input $2.5 Cached $0.25 Output $15	Input — Cached — Output —	Standard Long context after 272K input: input $5, cached $0.5, output $22.5.
gpt-audio	OpenAI	128K / 16K	Input $2.5 Cached Not offered Output $10	Input — Cached — Output —	Standard
gpt-audio-1.5	OpenAI	—	Input $2.5 Cached Not offered Output $10	Input — Cached — Output —	Standard
gpt-audio-mini	OpenAI	128K / 16K	Input $0.6 Cached Not offered Output $2.4	Input — Cached — Output —	Standard
gpt-image-1	OpenAI	—	Input $5 Cached $1.25 Output Not offered	Input — Cached — Output —	Standard
gpt-image-1-mini	OpenAI	—	Input $2 Cached $0.2 Output Not offered	Input — Cached — Output —	Standard
gpt-image-1.5	OpenAI	—	Input $5 Cached $1.25 Output $10	Input — Cached — Output —	Standard
gpt-image-2	OpenAI	—	Input $5 Cached $1.25 Output Not offered	Input — Cached — Output —	Standard
gpt-realtime	OpenAI	32K / 4K	Input $4 Cached $0.4 Output $16	Input — Cached — Output —	Standard
gpt-realtime-1.5	OpenAI	—	Input $4 Cached $0.4 Output $16	Input — Cached — Output —	Standard
gpt-realtime-mini	OpenAI	32K / 4K	Input $0.6 Cached $0.06 Output $2.4	Input — Cached — Output —	Standard
grok-4-1-fast-non-reasoning	Perplexity	128K	Input $0.2 Cached $0.05 Output $0.5	Input — Cached — Output —	Standard
nemotron-3-super-120b-a12b	Perplexity	32K	Input $0.25 Cached Not offered Output $2.5	Input — Cached — Output —	Standard
o1	OpenAI	200K / 100K	Input $15 Cached $7.5 Output $60	Input — Cached — Output —	Standard
o1-mini	OpenAI	200K / 100K	Input $1.1 Cached $0.55 Output $4.4	Input — Cached — Output —	Standard
o1-pro	OpenAI	200K / 100K	Input $150 Cached Not offered Output $600	Input — Cached — Output —	Standard
o3	OpenAI	200K / 100K	Input $2 Cached $0.5 Output $8	Input — Cached — Output —	Standard
o3-deep-research	OpenAI	200K / 100K	Input $10 Cached $2.5 Output $40	Input — Cached — Output —	Standard
o3-mini	OpenAI	200K / 100K	Input $1.1 Cached $0.55 Output $4.4	Input — Cached — Output —	Standard
o3-pro	OpenAI	200K / 100K	Input $20 Cached Not offered Output $80	Input — Cached — Output —	Standard
o4-mini	OpenAI	200K / 100K	Input $1.1 Cached $0.275 Output $4.4	Input — Cached — Output —	Standard
o4-mini-2025-04-16	OpenAI	200K	Input $4 Cached $1 Output $16	Input — Cached — Output —	Standard
o4-mini-2025-04-16 with data sharing	OpenAI	200K	Input $2 Cached $0.5 Output $8	Input — Cached — Output —	Standard
o4-mini-deep-research	OpenAI	200K / 100K	Input $2 Cached $0.5 Output $8	Input — Cached — Output —	Standard
Sonar	Perplexity	—	Input $1 Cached Not offered Output $1	Input — Cached — Output —	Standard
Sonar (LLM)	Perplexity	128K	Input $0.25 Cached $0.0625 Output $2.5	Input — Cached — Output —	Standard
Sonar Deep Research	Perplexity	—	Input $2 Cached Not offered Output $8	Input — Cached — Output —	Standard
Sonar Pro	Perplexity	—	Input $3 Cached Not offered Output $15	Input — Cached — Output —	Standard
Sonar Reasoning Pro	Perplexity	—	Input $2 Cached Not offered Output $8	Input — Cached — Output —	Standard
No models match your search or provider filter.

What the AI Token Calculator does

The AI Token Calculator is a free, privacy-first tool that tells you exactly how many tokens a piece of text, a source file or an image will consume when sent to a large language model (LLM). Tokens are the currency of modern AI: every request to GPT-4o, GPT-4, Claude, Gemini, Llama or any other model is measured, billed and capped in tokens rather than words. If you have ever been surprised by an API invoice, hit a mysterious “context length exceeded” error, or tried to squeeze a long document into a prompt, this calculator gives you the precise numbers you need before you spend a cent. Alongside the token total it also reports characters, characters without spaces, words, lines and sentences, so you get a complete picture of your content in one place.

How to use it

Getting a count takes just a few seconds:

Add your content. Paste or type directly into the text box, click Upload file to pick a document, or simply drag and drop a file or image anywhere onto the input box.
Choose a tokenizer that matches the model you are targeting — for example o200k_base for GPT-4o and the o-series, or cl100k_base for GPT-4 and GPT-3.5.
Read your results instantly. The token count and every other metric update live as you type or as soon as a file finishes loading — there is no “calculate” button to press.
Press Clear when you are done to wipe the content from memory. Nothing persists between sessions and nothing is ever saved.

Because everything runs locally, you can safely paste confidential prompts, proprietary code, contracts or customer data without it ever leaving your machine.

How it works

Language models do not read raw characters — they read tokens produced by a tokenizer. Most modern models use a technique called byte-pair encoding (BPE), which starts from individual bytes and repeatedly merges the most frequent adjacent pairs into larger units. The result is a vocabulary in which common words become a single token, rarer words split into several sub-word pieces, and spaces and punctuation are represented explicitly. This is why “dog” might be one token while “antidisestablishmentarianism” becomes many, and why the same word can tokenize differently depending on whether it is preceded by a space.

This calculator loads OpenAI’s actual open-source tokenizers in your browser and runs them over your text, so the count for OpenAI models is exact rather than estimated. When you pick the generic option, or if the tokenizer library cannot be loaded, it falls back to the well-known heuristic of roughly four characters per token and marks the figure with a “≈” to show it is an approximation. All of this happens client-side in JavaScript; your data never touches a server, which is what makes the tool both fast and private.

The tokenizer options explained

o200k_base — the newest OpenAI tokenizer, used by GPT-4o, GPT-4o mini, GPT-4.1 and the o1 / o3 reasoning models. It has a ~200k-token vocabulary and is more efficient on code and non-English text, so the same passage often uses fewer tokens than with older encoders.
cl100k_base — the tokenizer behind GPT-4, GPT-4 Turbo, GPT-3.5 Turbo and the text-embedding-3 models. Choose this whenever you are budgeting for those specific models.
Generic estimate (~4 chars/token) — a fast, model-agnostic approximation for providers whose tokenizers are not bundled here, such as Anthropic Claude, Google Gemini or Meta Llama. Treat it as a close ballpark; real counts on those platforms will be within a small margin.

Supported file types

You can drag in or upload files up to 10 MB. There are two categories:

Text & code (counted as text tokens):

.txt, .text, .md, .markdown, .mdx, .rst, .csv, .tsv, .json, .jsonl, .ndjson, .yaml, .yml, .toml, .ini, .env, .xml, .html, .htm, .css, .scss, .js, .mjs, .cjs, .jsx, .ts, .tsx, .py, .rb, .php, .java, .kt, .c, .h, .cpp, .cc, .cs, .go, .rs, .swift, .sh, .bash, .zsh, .sql, .log, .tex, .srt, .vtt — and most other UTF-8 text formats.

Images (measured as vision tokens):

.png, .jpg, .jpeg, .webp, .gif, .bmp

Binary formats such as PDF, DOCX, XLSX or ZIP are not decoded here because their tokenization depends on how a given model extracts text from them; export or copy the underlying text first for an accurate count.

How image (vision) tokens are counted

Multimodal models like GPT-4o charge for images in tokens too. When you drop in a PNG, JPG, WebP or GIF, the calculator reads the image’s pixel dimensions locally and applies OpenAI’s high-detail tiling formula: the image is first scaled to fit inside a 2048×2048 box, then its shortest side is scaled to 768 pixels, and the result is divided into 512-pixel tiles. Each tile costs 170 tokens on top of an 85-token base, so a small image might be a few hundred tokens while a large, detailed one can be well over a thousand. Low-detail mode, by contrast, is a flat 85 tokens regardless of size. The image itself is never uploaded — only its width and height are used.

How it helps you

Estimate cost before you call the API. Multiply your input and expected output tokens by a model’s per-token price — you can compare those prices across providers on our models directory — to forecast spend accurately.
Stay inside the context window. Every model has a maximum number of tokens it can handle in one request. Counting first prevents truncated prompts and “context length exceeded” errors.
Optimise prompts and RAG chunks. See instantly how trimming instructions, compressing examples or resizing retrieved passages changes your token footprint.
Budget fine-tuning and embeddings. Training files and embedding jobs are billed per token, so a quick count keeps large datasets predictable.
Protect sensitive data. Because the tool is fully client-side, you can measure private or regulated content without any compliance risk.

Tokens vs. words vs. characters

A common rule of thumb for English is that one token is about four characters or roughly 0.75 words, so 1,000 tokens is around 750 words and a single page of prose is often 500–650 tokens. These ratios shift with language, formatting and code: dense punctuation, emoji, non-Latin scripts and long identifiers all push the token count up relative to the word count. That is exactly why an exact tokenizer beats guessing — and why this page shows characters, words and tokens side by side so you can see the relationship for your own content.

Frequently asked questions

What is a token in AI models?

A token is the basic unit a language model reads and generates. Tokens are chunks of text produced by a tokenizer — often a whole short word, a word fragment, a space, or a punctuation mark. As a rough guide, one English token is about four characters or roughly three-quarters of a word, so 1,000 tokens is around 750 words. Models bill you and enforce context limits in tokens, not words, which is why counting them accurately matters.

How accurate is this token counter?

For OpenAI models the count is exact: we run the same byte-pair-encoding (BPE) tokenizers OpenAI uses — o200k_base for GPT-4o, GPT-4.1 and the o-series, and cl100k_base for GPT-4, GPT-3.5 Turbo and the v3 embeddings. For other providers such as Anthropic, Google or Meta, tokenization differs slightly, so use the generic estimate as a close approximation.

Is my text or file uploaded anywhere?

No. All counting happens entirely inside your browser using JavaScript. Your pasted text, uploaded documents and images are never transmitted to our servers, never logged, and never stored. Closing the tab or pressing Clear removes everything from memory.

Which file types are supported?

Plain-text and code files (.txt, .md, .csv, .json, .yaml, .xml, .html, .js, .ts, .py, .java, .go, .rb, .php, .sql, .log and many more) are counted as text. Image files (.png, .jpg, .jpeg, .webp, .gif) are measured as vision tokens using the image-tiling formula that GPT-4o-class models use.

How are image (vision) tokens calculated?

For images we estimate high-detail vision tokens the way OpenAI GPT-4o does: the image is scaled to fit within 2048×2048, then its shortest side is scaled to 768 pixels, and the result is divided into 512-pixel tiles. Each tile costs 170 tokens plus an 85-token base. Low-detail mode is a flat 85 tokens regardless of size.