Llama 4 token counter
Updated
In short: This counter runs Meta's official Llama 4 tokenizer.json in your browser, pinned to the same commit as our tokenizer dataset. It has 201,135 token IDs and is a 28 MB download. English prose takes about 126 tokens per 100 words, Modern Standard Arabic about 219.
tokens
| Model | Tokens | Chars per token | As input | As output |
|---|---|---|---|---|
| Llama 4 Scout 17B 16E Instruct Llama 4 tokenizer (Llama 4 Maverick) Price: Llama 4 Scout 17B 16E Instruct at GMI Cloud, cheapest listed input price (not BroadNet) | … | … | … | … |
| DeepSeek V4.1 Flash DeepSeek V4.1 Flash tokenizer | … | … | … | … |
| GLM 5.3 GLM 5.3 tokenizer | … | … | … | … |
| Qwen3.8 27B Qwen3.8 27B tokenizer | … | … | … | … |
| Gemma 4 VL 26B Gemma 4 tokenizer (Google base model) | … | … | … | … |
| GPT (o200k) OpenAI o200k_base | … | … | … | … |
| Llama 3.1 8B Instruct Llama 3 tokenizer (Llama 3.3 70B Instruct) Price: Llama 3.1 8B Instruct at DeepInfra, cheapest listed input price (not BroadNet) | … | … | … | … |
Cost = tokens × the model's published price per 1M input or output tokens. GPT is shown for reference only (no BroadNet price). Llama 4 and Llama 3 models are not sold by BroadNet: their rows use the cheapest standard input price we hold for a model with that tokenizer, read from the provider's own pricing page. Sources and check dates are in the real cost by language dataset. The chat-format option adds template tokens only for the DeepSeek, GLM, Qwen, Gemma and GPT rows; Llama 4 and Llama 3 rows count your text alone. Measured ratios for every tokenizer are in the tokenizer efficiency dataset. Rows marked ≈ are estimates from measured characters per token until you load that tokenizer; loading downloads its tokenizer.json from Hugging Face once and keeps it in your browser cache.
Other counters: DeepSeek · GLM · Qwen · Gemma · GPT · Llama 3 · Kimi · Mistral Small · Ministral 3 · gpt-oss · Phi-4 · Gemma 3 · MiniMax · Nemotron 3 · Qwen3 · all models. BroadNet prices synced 2026-11-14.
Report a wrong price or number (email to info@broadnet.ai, with this page's link and your inputs filled in).
How it works
- Paste text into the box or pick a sample. The Llama 4 tokenizer (Llama 4 Maverick) is preselected and downloads once (28 MB) when you start typing.
- The text appears split into tokens, a colour per token; tick "Show token IDs" to see the numbers.
- Compare the Llama 4 count with the other tokenizers in the table; load any of them for an exact count.
- Copy the link or download a CSV of the counts and costs.
About the Llama 4 tokenizer
The Llama 4 tokenizer is a byte-level BPE. It has 201,135 token IDs, 1,135 of them added or special tokens such as chat-format markers. Text is split into words, punctuation and numbers first, then merged byte by byte. Numbers are split into groups of up to three digits, so 2026 is two tokens.
On the open texts of our tokenizer efficiency dataset it needs 1,264 tokens per 1,000 English words (rank 2 of the 16 tokenizers we measured, 1 = fewest), 2,025 per 1,000 words of Python (rank 7 of 16) and 2,186 per 1,000 words of Modern Standard Arabic (rank 10 of 16). Its best showing here is JSON (rank 1 of 16); its weakest is Arabic MSA (rank 10 of 16).
2 models in our price database use this tokenizer: Llama 4 Maverick 17B 128E Instruct FP8 and Llama 4 Scout 17B 16E Instruct. We hold 13 standard prices for them from 9 providers. The cheapest input price is $0.08 per 1M tokens, for Llama 4 Scout 17B 16E Instruct on GMI Cloud (checked 2026-11-15). At that price 1,000 English words cost $0.00010 to send. Llama 4 models are not sold by BroadNet. That is less than BroadNet's cheapest chat model, Gemma 4 VL 26B, at $0.00013 for the same text.
| Tokenizer | Llama 4 tokenizer (Llama 4 Maverick), a byte-level BPE |
|---|---|
| Source | meta-llama/Llama-4-Maverick-17B-128E-Instruct (commit 73d1471), read from unsloth/Llama-4-Maverick-17B-128E-Instruct (byte-identical) |
| Token IDs | 201,135 (200,000 in the base vocabulary) |
| Models using it | 2: Llama 4 Maverick 17B 128E Instruct FP8 and Llama 4 Scout 17B 16E Instruct |
| Cheapest input price | $0.08 per 1M tokens (Llama 4 Scout 17B 16E Instruct, GMI Cloud, checked 2026-11-15) |
| English prose | 126 tokens per 100 words (rank 2 of 16) |
| Arabic (MSA) | 219 tokens per 100 words (rank 10 of 16) |
| Python code | 203 tokens per 100 words (rank 7 of 16) |
| Download | 28 MB, once; then cached in your browser |
Tokens per 100 words measured on the open texts of our dataset with all 16 tokenizers (2026-11-15); rank 1 = fewest tokens on the same texts. Model mapping checked 2026-11-15. Third-party prices are standard-tier prices from each provider's own page; BroadNet prices come from its pricing.json.
How much does a Llama 4 prompt cost?
input cost = tokens × input price per 1M ÷ 1,000,000
output cost = tokens × output price per 1M ÷ 1,000,000Worked example: "English Wikipedia: Water" in our dataset is 274 words. The Llama 4 tokenizer (Llama 4 Maverick) splits it into 348 tokens. Sent to Llama 4 Scout 17B 16E Instruct on GMI Cloud at $0.08 per 1M input tokens it costs $0.000028; the same number of tokens written as output at $0.50 per 1M costs $0.00017.
Questions
Which tokenizer does Llama 4 use?
A byte-level BPE with 201,135 token IDs, published as tokenizer.json in meta-llama/Llama-4-Maverick-17B-128E-Instruct. This page loads that file pinned to commit 73d1471, the commit our dataset measured. The official repo is gated (it needs a Hugging Face account and licence acceptance), so the file is read from unsloth/Llama-4-Maverick-17B-128E-Instruct. Our dataset checks that the mirror's file is byte-identical to the official one: same SHA-256 and the same Git object ID that the official repo lists.
Which models use the Llama 4 tokenizer?
Llama 4 Maverick 17B 128E Instruct FP8 and Llama 4 Scout 17B 16E Instruct. Each was checked against the model's own Hugging Face repo on 2026-11-15: same tokenizer file, or the same vocabulary and rules with identical counts on every dataset text.
What is the cheapest Llama 4 API price per token?
Of the 13 standard prices we hold for Llama 4 models, the lowest input price is $0.08 per 1M tokens (output $0.50), for Llama 4 Scout 17B 16E Instruct on GMI Cloud, read from the provider's pricing page on 2026-11-15. Promotional prices are not ranked. Prices differ by model size and provider: each model's price page lists every provider we hold.
How many tokens does Arabic take with Llama 4?
Modern Standard Arabic sentences take 1.59× the tokens of their English translations with this tokenizer (rank 10 of 16 on the same texts). Hindi takes 1.65× and Chinese 1.07×. Measured on parallel sentences in our open dataset (2026-11-15). Paste your own text above, or use the Arabic token counter.
Is the Llama 4 tokenizer the same as Llama 3?
No. On the 52 texts of our dataset they gave the same count on 2; in total Llama 4 needs 16% fewer tokens, and up to 44% on a single text. See the Llama 3 counter.
Does the count include chat-template tokens?
No. For Llama 4 this page counts your text only, as our dataset does. A chat request also adds the chat template's role markers and special tokens, so the billed prompt is a little longer than your text. Read usage.prompt_tokens on a real response to see the exact figure.
Go deeper
Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity
Sources
- BroadNet Inference machine-readable prices (pricing.json)
- meta-llama/Llama-4-Maverick-17B-128E-Instruct on Hugging Face (tokenizer.json, commit 73d1471)
- unsloth/Llama-4-Maverick-17B-128E-Instruct: byte-identical ungated copy (commit 86c5ecb)
- GMI Cloud pricing for Llama 4 Scout 17B 16E Instruct (checked 2026-11-15)