AI Tokenizer & Token Cost Calculator
Count tokens with the exact OpenAI BPE encodings, see where every token boundary falls, and price your prompt across 95 live models — all computed in your browser.
- No cloud storageYour prompts never leave the browser
- Client-side BPEExact o200k & cl100k token boundaries
- No account requiredFree forever, no paywall, no ads
0 / 1,000,000 tokens0.00%
1,000,000 tokens remaining before this model's limit.
Safe range
No Tokens Generated
Type or paste text in the raw editor to inspect individual BPE token boundaries, hex encodings, and byte weights in real-time.
Workload Budget Calculator
Model what this prompt actually costs to run at volume. Set your batch size and expected output length, then compare standard input pricing against cached input pricing across every model in the live catalog. Models without a cache tier are marked — rather than given a fabricated discount.
| Model | Standard Input | Cached Prompt | Single Run | Batch Total (1,000) |
|---|---|---|---|---|
Claude Sonnet 5.5 Anthropic • Claude BPE | $2.000 / 1M | $0.100 / 1M-95% | $0.00000 | $0.00 |
Claude Haiku 5.5TIERS Anthropic • Claude BPE | $0.100 / 1M | $0.010 / 1M-90% | $0.00000 | $0.00 |
Claude Opus 5 Anthropic • Claude BPE | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
Claude Sonnet 5ACTIVE Anthropic • Claude BPE | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT-6.1 Sol ProTIERS OpenAI • o200k_base | $2.000 / 1M | $0.100 / 1M-95% | $0.00000 | $0.00 |
GPT-6.1 SolTIERS OpenAI • o200k_base | $2.000 / 1M | $0.100 / 1M-95% | $0.00000 | $0.00 |
GPT-6 LunaTIERS OpenAI • o200k_base | $0.100 / 1M | $0.010 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 SolTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 TerraTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 LunaTIERS OpenAI • o200k_base | $0.200 / 1M | $0.020 / 1M-90% | $0.00000 | $0.00 |
Gemini 3.8 Flash Google • Gemini | $0.750 / 1M | $0.075 / 1M-90% | $0.00000 | $0.00 |
Gemini 3.7 Flash Google • Gemini | $0.750 / 1M | $0.075 / 1M-90% | $0.00000 | $0.00 |
Gemini 3.5 Flash Lite Google • Gemini | $0.300 / 1M | $0.030 / 1M-90% | $0.00000 | $0.00 |
Llama 4 Maverick Meta • Llama | $0.188 / 1M | $0.050 / 1M-73% | $0.00000 | $0.00 |
DeepSeek V4.1 Flash DeepSeek • DeepSeek | $0.300 / 1M | $0.006 / 1M-98% | $0.00000 | $0.00 |
DeepSeek V4 Pro 0813TIERS DeepSeek • DeepSeek | $0.660 / 1M | $0.022 / 1M-97% | $0.00000 | $0.00 |
DeepSeek V4 Flash 0731 DeepSeek • DeepSeek | $0.010 / 1M | $0.010 / 1M-0% | $0.00000 | $0.00 |
Grok 4.6TIERS xAI • Grok | $2.000 / 1M | $0.500 / 1M-75% | $0.00000 | $0.00 |
Qwen3.8 Max (0902) Alibaba • Qwen | $2.000 / 1M | $0.250 / 1M-88% | $0.00000 | $0.00 |
Qwen3.7 FlashTIERS Alibaba • Qwen | $0.030 / 1M | $0.006 / 1M-80% | $0.00000 | $0.00 |
Mistral Large 4 Mistral • Mistral | $0.680 / 1M | $0.070 / 1M-90% | $0.00000 | $0.00 |
GPT-4 OpenAI • cl100k_base | $30.000 / 1M | — | $0.00000 | $0.00 |
GPT-3.5 Turbo OpenAI • cl100k_base | $0.500 / 1M | — | $0.00000 | $0.00 |
Claude Opus 5.5 Anthropic • Claude BPE | $4.000 / 1M | $0.200 / 1M-95% | $0.00000 | $0.00 |
DeepSeek V4 Flash 0423 DeepSeek • DeepSeek | $0.010 / 1M | $0.010 / 1M-0% | $0.00000 | $0.00 |
GPT-6 AstraTIERS OpenAI • o200k_base | $10.000 / 1M | $1.000 / 1M-90% | $0.00000 | $0.00 |
GPT-6 SolTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
Gemini 3 Flash Preview Google • Gemini | $0.500 / 1M | $0.050 / 1M-90% | $0.00000 | $0.00 |
DeepSeek V4 Pro 0423 DeepSeek • DeepSeek | $0.955 / 1M | $0.080 / 1M-92% | $0.00000 | $0.00 |
Gemini 2.5 Flash Lite Google • Gemini | $0.100 / 1M | $0.010 / 1M-90% | $0.00000 | $0.00 |
GPT-6 Luna ProTIERS OpenAI • o200k_base | $0.100 / 1M | $0.010 / 1M-90% | $0.00000 | $0.00 |
Qwen3.8 27B Alibaba • Qwen | $0.425 / 1M | $0.085 / 1M-80% | $0.00000 | $0.00 |
Grok 4.7TIERS xAI • Grok | $2.000 / 1M | $0.500 / 1M-75% | $0.00000 | $0.00 |
gpt-oss-120b OpenAI • o200k_base | $0.037 / 1M | — | $0.00000 | $0.00 |
Qwen3.8 Flash Alibaba • Qwen | $0.150 / 1M | $0.016 / 1M-89% | $0.00000 | $0.00 |
Gemini 3.1 Flash Lite Google • Gemini | $0.250 / 1M | $0.025 / 1M-90% | $0.00000 | $0.00 |
Gemini 2.5 Flash Google • Gemini | $0.300 / 1M | $0.030 / 1M-90% | $0.00000 | $0.00 |
Gemma 4 31B Google • Gemma | $0.090 / 1M | $0.050 / 1M-44% | $0.00000 | $0.00 |
DeepSeek V3.2 DeepSeek • DeepSeek | $0.259 / 1M | $0.135 / 1M-48% | $0.00000 | $0.00 |
Gemma 4 26B A4B Google • Gemma | $0.090 / 1M | $0.050 / 1M-44% | $0.00000 | $0.00 |
Claude Sonnet 4.6 Anthropic • Claude BPE | $3.000 / 1M | $0.300 / 1M-90% | $0.00000 | $0.00 |
Claude Fable 5.1 Anthropic • Claude BPE | $10.000 / 1M | $0.250 / 1M-98% | $0.00000 | $0.00 |
Claude Haiku 4.5 Anthropic • Claude BPE | $1.000 / 1M | $0.100 / 1M-90% | $0.00000 | $0.00 |
gpt-oss-20b OpenAI • o200k_base | $0.018 / 1M | $0.009 / 1M-50% | $0.00000 | $0.00 |
Gemini 3.1 Pro PreviewTIERS Google • Gemini | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
Nano Banana 2.1 Google • Gemini | $1.500 / 1M | — | $0.00000 | $0.00 |
Qwen3.8 Max Prime Alibaba • Qwen | $4.000 / 1M | $0.500 / 1M-88% | $0.00000 | $0.00 |
GPT-6 Sol ProTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT-6 Astra ProTIERS OpenAI • o200k_base | $10.000 / 1M | $1.000 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 Luna ProTIERS OpenAI • o200k_base | $0.200 / 1M | $0.020 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 Terra ProTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT-5.6 Sol ProTIERS OpenAI • o200k_base | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
GPT Chat Latest OpenAI • o200k_base | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
GPT-5.5 ProTIERS OpenAI • o200k_base | $30.000 / 1M | — | $0.00000 | $0.00 |
Claude Fable 5 Anthropic • Claude BPE | $10.000 / 1M | $1.000 / 1M-90% | $0.00000 | $0.00 |
Claude Opus 4.8 Anthropic • Claude BPE | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
Claude Opus 4.7 Anthropic • Claude BPE | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
Claude Opus 4.6 Anthropic • Claude BPE | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
Claude Opus 4.5 Anthropic • Claude BPE | $5.000 / 1M | $0.500 / 1M-90% | $0.00000 | $0.00 |
Claude Sonnet 4.5TIERS Anthropic • Claude BPE | $3.000 / 1M | $0.300 / 1M-90% | $0.00000 | $0.00 |
Gemini 3.6 Flash Google • Gemini | $0.750 / 1M | $0.075 / 1M-90% | $0.00000 | $0.00 |
Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) Google • Gemini | $0.250 / 1M | — | $0.00000 | $0.00 |
Nano Banana 2 (Gemini 3.1 Flash Image) Google • Gemini | $0.500 / 1M | — | $0.00000 | $0.00 |
Nano Banana Pro (Gemini 3 Pro Image) Google • Gemini | $2.000 / 1M | $0.200 / 1M-90% | $0.00000 | $0.00 |
Gemini 3.5 Flash Google • Gemini | $1.500 / 1M | $0.150 / 1M-90% | $0.00000 | $0.00 |
Lyria 3 Pro Preview Google • Other BPE | $0.000 / 1M | — | $0.00000 | $0.00 |
DeepSeek V4 Flash Vision Exp DeepSeek • DeepSeek | $0.216 / 1M | $0.007 / 1M-97% | $0.00000 | $0.00 |
DeepSeek V3.2 Exp DeepSeek • DeepSeek | $0.270 / 1M | — | $0.00000 | $0.00 |
DeepSeek V3.1 Terminus DeepSeek • DeepSeek | $0.270 / 1M | — | $0.00000 | $0.00 |
DeepSeek V3.1 DeepSeek • DeepSeek | $0.250 / 1M | $0.130 / 1M-48% | $0.00000 | $0.00 |
R1 0528 DeepSeek • DeepSeek | $0.500 / 1M | $0.350 / 1M-30% | $0.00000 | $0.00 |
DeepSeek V3 0324 DeepSeek • DeepSeek | $0.290 / 1M | $0.110 / 1M-62% | $0.00000 | $0.00 |
Llama Guard 4 12B Meta • Other BPE | $0.180 / 1M | — | $0.00000 | $0.00 |
Llama 4 Scout Meta • Llama | $0.100 / 1M | — | $0.00000 | $0.00 |
Llama 3.3 70B Instruct Meta • Llama | $0.100 / 1M | — | $0.00000 | $0.00 |
Llama 3.2 1B Instruct Meta • Llama | $0.027 / 1M | — | $0.00000 | $0.00 |
Llama 3.2 3B Instruct Meta • Llama | $0.050 / 1M | — | $0.00000 | $0.00 |
Llama 3.1 70B Instruct Meta • Llama | $0.400 / 1M | — | $0.00000 | $0.00 |
Mistral Medium 3.5 Mistral • Mistral | $1.500 / 1M | — | $0.00000 | $0.00 |
Mistral Small 4 Mistral • Mistral | $0.150 / 1M | $0.015 / 1M-90% | $0.00000 | $0.00 |
Devstral 2 2512 Mistral • Mistral | $0.400 / 1M | $0.040 / 1M-90% | $0.00000 | $0.00 |
Ministral 3 14B 2512 Mistral • Mistral | $0.200 / 1M | $0.020 / 1M-90% | $0.00000 | $0.00 |
Ministral 3 8B 2512 Mistral • Mistral | $0.150 / 1M | $0.015 / 1M-90% | $0.00000 | $0.00 |
Ministral 3 3B 2512 Mistral • Mistral | $0.100 / 1M | $0.010 / 1M-90% | $0.00000 | $0.00 |
Grok 4.5TIERS xAI • Grok | $2.000 / 1M | $0.300 / 1M-85% | $0.00000 | $0.00 |
Grok Build 0.1TIERS xAI • Grok | $1.000 / 1M | $0.200 / 1M-80% | $0.00000 | $0.00 |
Grok 4.3TIERS xAI • Grok | $1.250 / 1M | $0.200 / 1M-84% | $0.00000 | $0.00 |
Grok 4.20 Multi-AgentTIERS xAI • Grok | $1.250 / 1M | $0.200 / 1M-84% | $0.00000 | $0.00 |
Grok 4.20TIERS xAI • Grok | $1.250 / 1M | $0.200 / 1M-84% | $0.00000 | $0.00 |
Qwen3.8 Omni Flash Alibaba • Qwen | $0.150 / 1M | $0.016 / 1M-89% | $0.00000 | $0.00 |
Qwen3.8 2.4T A95B Alibaba • Qwen | $2.000 / 1M | $0.250 / 1M-88% | $0.00000 | $0.00 |
Qwen3.7 PlusTIERS Alibaba • Qwen | $0.320 / 1M | $0.064 / 1M-80% | $0.00000 | $0.00 |
Qwen3.7 Max Alibaba • Qwen | $1.475 / 1M | $0.295 / 1M-80% | $0.00000 | $0.00 |
Qwen3.5 Plus 2026-04-20TIERS Alibaba • Qwen | $0.300 / 1M | — | $0.00000 | $0.00 |
Qwen3.6 FlashTIERS Alibaba • Qwen | $0.188 / 1M | — | $0.00000 | $0.00 |
How Tokenization Actually Works
The vocabulary behind every token count you see above — and why the same sentence costs more on one model than another.
What is an AI Tokenizer?
Large language models don't read words or characters directly; they process text as numeric IDs called tokens. A tokenizer is a deterministic algorithm that decomposes raw text into these tokens using a fixed dictionary. Tokens represent whole words, word fragments, punctuation, or individual bytes.
How are tokens counted?
Token counting applies Byte Pair Encoding (BPE) merge rules against text. Starting from individual bytes, the algorithm iteratively merges the most frequent pairs in its vocabulary into composite tokens. Formatting, spaces, capitalization, and numbers directly alter merge boundaries.
How many words are in 1,000 tokens?
For standard English prose, 1,000 tokens is approximately 750 words (~0.75 words per token, or ~1.33 tokens per word). For source code, JSON, and LaTeX, the ratio shifts closer to 0.5–0.6 words per token due to indentation, punctuation, and camelCase syntax.
How many tokens are in 1,000 words?
In general English, 1,000 words yields approximately 1,300 to 1,400 tokens. Older vocabularies like cl100k_base trend higher (~1,350 tokens), while modern o200k_base compresses this by 5–15% thanks to its expanded 200,000-token dictionary.
What is the difference between o200k and cl100k?
cl100k_base (100k vocab) was introduced with GPT-4 and remains the standard for vector embeddings (text-embedding-3). o200k_base (200k vocab) powers GPT-4o, o1, and GPT-6, compressing code by ~15% and non-Latin scripts (Devanagari, CJK) by up to 25–40%.
How does prompt caching reduce API costs?
Prompt caching allows providers (Anthropic, OpenAI, DeepSeek, Google) to reuse pre-computed KV-cache states for identical prompt prefixes. Reusing stable context discounts input pricing by 75% to 95% (e.g. Claude Sonnet 5.5 drops from $2.00 to $0.10/1M on cache hits).
Frequently Asked Questions
How TokenCraft handles your text, your data, and your money.