Token
counter.
Paste text, see how every major LLM tokenizes it — and what it costs as input or output on 124 live-priced models. OpenAI counted exactly (o200k_base); 9 more vendors calibrated on their own published tokenizers; the rest estimated and labeled honestly. Everything runs in your browser.
Jump to counter Private by design — text never leaves the page
Counted locally in your browser — text never leaves this page
Characters 482
Words 78
Tokens · range across vendors 90 – 185
MiniMax ≈90 cal Alibaba (Qwen) ≈92 cal Baichuan ≈92 cal DeepSeek ≈92 cal Tencent (Hunyuan) ≈92 cal Zhipu (Z.ai / GLM) ≈92 cal ByteDance (Doubao) ≈93 cal 01.AI ≈95 cal Mistral ≈95 cal OpenAI 102 exact xAI ≈115–125 est Google ≈125 est Moonshot (Kimi) ≈125 est Anthropic ≈138–185 est
Same text, different vendors — tokenizers disagree, so counts (and bills) do too. exact = o200k_base BPE · cal = calibrated on the vendor's own published tokenizer · est = editorial estimate (no public tokenizer).
All models
Your text on every model
Token count, share of the context window, and the cost of this exact text as input or as generated output — for all 124 live-priced models. Click a column to sort; click a model for its full pricing hub.
| Model | Tokens | Method | % of context | In $/1M | Out $/1M | As input | As output |
|---|---|---|---|---|---|---|---|
| Leanstral Mistral | 95 | ≈ cal | — | $0.00 | $0.00 | $0 | $0 |
| Hunyuan Lite Tencent (Hunyuan) | 92 | ≈ cal | — | $0.00 | $0.00 | $0 | $0 |
| GLM-4.5 Flash Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.00 | $0.00 | $0 | $0 |
| GLM-4.7 Flash Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.00 | $0.00 | $0 | $0 |
| Doubao Seed 1.6 Flash ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.02 | $0.21 | $0.00000 | $0.00002 |
| Doubao Seed 2.0 Mini ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.03 | $0.28 | $0.00000 | $0.00003 |
| Doubao Seed 1.6 Lite ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.04 | $0.34 | $0.00000 | $0.00003 |
| Qwen3 VL Flash Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.05 | $0.40 | $0.00000 | $0.00004 |
| Doubao Seed 1.6 Vision ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.06 | $0.56 | $0.00001 | $0.00005 |
| Hunyuan A13B Tencent (Hunyuan) | 92 | ≈ cal | <0.1% | $0.07 | $0.28 | $0.00001 | $0.00003 |
| GLM-4.7 FlashX Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.07 | $0.40 | $0.00001 | $0.00004 |
| Doubao Seed 2.0 Lite ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.09 | $0.51 | $0.00001 | $0.00005 |
| Qwen 3.5 Flash Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.10 | $0.40 | $0.00001 | $0.00004 |
| GLM-4 32B (0414, 128K) Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.10 | $0.10 | $0.00001 | $0.00001 |
| Devstral Small 2 Mistral | 95 | ≈ cal | — | $0.10 | $0.30 | $0.00001 | $0.00003 |
| Ministral 3 3B Mistral | 95 | ≈ cal | <0.1% | $0.10 | $0.10 | $0.00001 | $0.00001 |
| Mistral Small 3.2 Mistral | 95 | ≈ cal | — | $0.10 | $0.30 | $0.00001 | $0.00003 |
| Mistral Small 4 Mistral | 95 | ≈ cal | <0.1% | $0.10 | $0.30 | $0.00001 | $0.00003 |
| Hunyuan TurboS Tencent (Hunyuan) | 92 | ≈ cal | — | $0.11 | $0.28 | $0.00001 | $0.00003 |
| Doubao Seed 1.6 ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.11 | $1.13 | $0.00001 | $0.00010 |
| Doubao Seed 1.8 ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.11 | $1.13 | $0.00001 | $0.00010 |
| Doubao Seed Character ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.11 | $0.28 | $0.00001 | $0.00003 |
| Gemini 2.5 Flash-Lite Google | 125 | ≈ est | <0.1% | $0.10 | $0.40 | $0.00001 | $0.00005 |
| Baichuan4 Air Baichuan | 92 | ≈ cal | 0.3% | $0.14 | $0.14 | $0.00001 | $0.00001 |
| DeepSeek V4 Flash DeepSeek | 92 | ≈ cal | <0.1% | $0.14 | $0.28 | $0.00001 | $0.00003 |
| Hunyuan T1 Tencent (Hunyuan) | 92 | ≈ cal | — | $0.14 | $0.56 | $0.00001 | $0.00005 |
| Hunyuan Translation Lite Tencent (Hunyuan) | 92 | ≈ cal | — | $0.14 | $0.42 | $0.00001 | $0.00004 |
| Yi Lightning 01.AI | 95 | ≈ cal | 0.6% | $0.14 | $0.14 | $0.00001 | $0.00001 |
| Ministral 3 8B Mistral | 95 | ≈ cal | <0.1% | $0.15 | $0.15 | $0.00001 | $0.00001 |
| Qwen3 32B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.16 | $0.64 | $0.00001 | $0.00006 |
| Qwen3 Next 80B A3B Instruct Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.16 | $1.30 | $0.00001 | $0.00012 |
| Qwen3 Next 80B A3B Thinking Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.16 | $1.30 | $0.00001 | $0.00012 |
| Hunyuan Translation Tencent (Hunyuan) | 92 | ≈ cal | — | $0.17 | $0.51 | $0.00002 | $0.00005 |
| Doubao Seed Code ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.17 | $1.13 | $0.00002 | $0.00010 |
| Doubao Seed Translation ByteDance (Doubao) | 93 | ≈ cal | — | $0.17 | $0.51 | $0.00002 | $0.00005 |
| Qwen3 8B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.20 | $0.76 | $0.00002 | $0.00007 |
| Qwen3 VL Plus Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.20 | $1.60 | $0.00002 | $0.00015 |
| GLM-4.5 Air Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.20 | $1.10 | $0.00002 | $0.00010 |
| Ministral 3 14B Mistral | 95 | ≈ cal | <0.1% | $0.20 | $0.20 | $0.00002 | $0.00002 |
| Qwen3 30B A3B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.22 | $0.87 | $0.00002 | $0.00008 |
| Qwen3 30B A3B Instruct 2507 Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.22 | $0.87 | $0.00002 | $0.00008 |
| Qwen3 30B A3B Thinking 2507 Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.22 | $2.60 | $0.00002 | $0.00024 |
| GPT-5.4 nano OpenAI | 102 | exact | <0.1% | $0.20 | $1.25 | $0.00002 | $0.00013 |
| Qwen3 235B A22B Instruct 2507 Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.25 | $1.00 | $0.00002 | $0.00009 |
| Qwen3 235B A22B Thinking 2507 Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.25 | $2.49 | $0.00002 | $0.00023 |
| Baichuan-M2 Baichuan | 92 | ≈ cal | 0.3% | $0.28 | $2.82 | $0.00003 | $0.00026 |
| QwQ 32B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.29 | $0.86 | $0.00003 | $0.00008 |
| MiniMax M2 MiniMax | 90 | ≈ cal | <0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| MiniMax M2-her MiniMax | 90 | ≈ cal | 0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| MiniMax M2.1 MiniMax | 90 | ≈ cal | <0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| MiniMax M2.5 MiniMax | 90 | ≈ cal | <0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| MiniMax M2.7 MiniMax | 90 | ≈ cal | <0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| MiniMax M3 MiniMax | 90 | ≈ cal | <0.1% | $0.30 | $1.20 | $0.00003 | $0.00011 |
| Qwen3 Coder Flash Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.30 | $1.50 | $0.00003 | $0.00014 |
| Codestral Mistral | 95 | ≈ cal | — | $0.30 | $0.90 | $0.00003 | $0.00009 |
| Gemini 3.1 Flash-Lite Google | 125 | ≈ est | <0.1% | $0.25 | $1.50 | $0.00003 | $0.00019 |
| Qwen3 14B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.35 | $1.40 | $0.00003 | $0.00013 |
| Qwen 3.5 122B A10B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.40 | $3.20 | $0.00004 | $0.00029 |
| Qwen 3.5 Plus Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.40 | $2.40 | $0.00004 | $0.00022 |
| Gemini 2.5 Flash Google | 125 | ≈ est | <0.1% | $0.30 | $2.50 | $0.00004 | $0.00031 |
| Devstral 2 Mistral | 95 | ≈ cal | — | $0.40 | $2.00 | $0.00004 | $0.00019 |
| Hunyuan T1 Vision Tencent (Hunyuan) | 92 | ≈ cal | 0.3% | $0.42 | $1.27 | $0.00004 | $0.00012 |
| Hunyuan TurboS Vision Tencent (Hunyuan) | 92 | ≈ cal | 0.3% | $0.42 | $1.27 | $0.00004 | $0.00012 |
| Hunyuan TurboS Vision Video Tencent (Hunyuan) | 92 | ≈ cal | 0.4% | $0.42 | $1.27 | $0.00004 | $0.00012 |
| Tencent HY Vision 1.5 Instruct Tencent (Hunyuan) | 92 | ≈ cal | 0.4% | $0.42 | $1.27 | $0.00004 | $0.00012 |
| DeepSeek V4 Pro DeepSeek | 92 | ≈ cal | <0.1% | $0.43 | $0.87 | $0.00004 | $0.00008 |
| Qwen3.7 Plus Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.44 | $1.77 | $0.00004 | $0.00016 |
| Hunyuan 2.0 Instruct Tencent (Hunyuan) | 92 | ≈ cal | <0.1% | $0.45 | $1.12 | $0.00004 | $0.00010 |
| Doubao Seed 2.0 Code ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.45 | $2.25 | $0.00004 | $0.00021 |
| Doubao Seed 2.0 Pro ByteDance (Doubao) | 93 | ≈ cal | <0.1% | $0.45 | $2.25 | $0.00004 | $0.00021 |
| Magistral Small Mistral | 95 | ≈ cal | — | $0.50 | $1.50 | $0.00005 | $0.00014 |
| Mistral Large 3 Mistral | 95 | ≈ cal | — | $0.50 | $1.50 | $0.00005 | $0.00014 |
| Hunyuan 2.0 Think (HYThink) Tencent (Hunyuan) | 92 | ≈ cal | <0.1% | $0.56 | $2.24 | $0.00005 | $0.00021 |
| MiniMax M2.1 Highspeed MiniMax | 90 | ≈ cal | <0.1% | $0.60 | $2.40 | $0.00005 | $0.00022 |
| MiniMax M2.5 Highspeed MiniMax | 90 | ≈ cal | <0.1% | $0.60 | $2.40 | $0.00005 | $0.00022 |
| MiniMax M2.7 Highspeed MiniMax | 90 | ≈ cal | <0.1% | $0.60 | $2.40 | $0.00005 | $0.00022 |
| Qwen 3.5 397B A17B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.60 | $3.60 | $0.00006 | $0.00033 |
| GLM-4.5 Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.60 | $2.20 | $0.00006 | $0.00020 |
| GLM-4.6 Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.60 | $2.20 | $0.00006 | $0.00020 |
| GLM-4.7 Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $0.60 | $2.20 | $0.00006 | $0.00020 |
| Qwen3 235B A22B Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.70 | $2.80 | $0.00006 | $0.00026 |
| Baichuan-M3-Plus Baichuan | 92 | ≈ cal | 0.3% | $0.70 | $1.27 | $0.00006 | $0.00012 |
| QwQ Plus Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $0.80 | $2.40 | $0.00007 | $0.00022 |
| Kimi K2.5 Moonshot (Kimi) | 125 | ≈ est | <0.1% | $0.60 | $3.00 | $0.00007 | $0.00038 |
| GPT-5.4 mini OpenAI | 102 | exact | <0.1% | $0.75 | $4.50 | $0.00008 | $0.00046 |
| Qwen3 Coder Plus Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $1.00 | $5.00 | $0.00009 | $0.00046 |
| GLM-5 Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $1.00 | $3.20 | $0.00009 | $0.00029 |
| GLM-4.5 AirX Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $1.10 | $4.50 | $0.00010 | $0.00041 |
| Qwen3 Max Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $1.20 | $6.00 | $0.00011 | $0.00055 |
| GLM-5 Turbo Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $1.20 | $4.00 | $0.00011 | $0.00037 |
| Kimi K2.6 Moonshot (Kimi) | 125 | ≈ est | <0.1% | $0.95 | $4.00 | $0.00012 | $0.00050 |
| GLM-5.1 Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $1.40 | $4.40 | $0.00013 | $0.00040 |
| Baichuan-M2-Plus Baichuan | 92 | ≈ cal | 0.3% | $1.41 | $4.22 | $0.00013 | $0.00039 |
| Baichuan-M3 Baichuan | 92 | ≈ cal | 0.3% | $1.41 | $4.22 | $0.00013 | $0.00039 |
| Claude Haiku 4.5 Anthropic | 138 | ≈ est | <0.1% | $1.00 | $5.00 | $0.00014 | $0.00069 |
| Mistral Medium 3.1 Mistral | 95 | ≈ cal | — | $1.50 | $7.50 | $0.00014 | $0.00071 |
| Mistral Medium 3.5 Mistral | 95 | ≈ cal | — | $1.50 | $7.50 | $0.00014 | $0.00071 |
| Grok 4.20 (0309) Non-Reasoning xAI | 115 | ≈ est | <0.1% | $1.25 | $2.50 | $0.00014 | $0.00029 |
| Grok 4.20 (0309) Reasoning xAI | 115 | ≈ est | <0.1% | $1.25 | $2.50 | $0.00014 | $0.00029 |
| Grok 4.20 Multi-Agent (0309) xAI | 115 | ≈ est | <0.1% | $1.25 | $2.50 | $0.00014 | $0.00029 |
| Baichuan3-Turbo Baichuan | 92 | ≈ cal | 0.3% | $1.69 | $1.69 | $0.00016 | $0.00016 |
| Gemini 2.5 Pro Google | 125 | ≈ est | <0.1% | $1.25 | $10.0 | $0.00016 | $0.00125 |
| Grok 4.3 xAI | 125 | ≈ est | <0.1% | $1.25 | $2.50 | $0.00016 | $0.00031 |
| GPT-5.3-Codex OpenAI | 102 | exact | <0.1% | $1.75 | $14.0 | $0.00018 | $0.00143 |
| Gemini 3.5 Flash Google | 125 | ≈ est | <0.1% | $1.50 | $9.00 | $0.00019 | $0.00113 |
| Magistral Medium Mistral | 95 | ≈ cal | — | $2.00 | $5.00 | $0.00019 | $0.00048 |
| Baichuan4 Turbo Baichuan | 92 | ≈ cal | 0.3% | $2.11 | $2.11 | $0.00019 | $0.00019 |
| GLM-4.5 X Zhipu (Z.ai / GLM) | 92 | ≈ cal | <0.1% | $2.20 | $8.90 | $0.00020 | $0.00082 |
| Qwen3.7 Max Alibaba (Qwen) | 92 | ≈ cal | <0.1% | $2.77 | $8.31 | $0.00025 | $0.00076 |
| GPT-5.4 OpenAI | 102 | exact | <0.1% | $2.50 | $15.0 | $0.00026 | $0.00153 |
| Baichuan3-Turbo (128K) Baichuan | 92 | ≈ cal | <0.1% | $3.38 | $3.38 | $0.00031 | $0.00031 |
| Claude Sonnet 4.5 Anthropic | 138 | ≈ est | <0.1% | $3.00 | $15.0 | $0.00041 | $0.00207 |
| Claude Sonnet 4.6 Anthropic | 138 | ≈ est | <0.1% | $3.00 | $15.0 | $0.00041 | $0.00207 |
| chat-latest OpenAI | 102 | exact | <0.1% | $5.00 | $30.0 | $0.00051 | $0.00306 |
| GPT-5.5 OpenAI | 102 | exact | <0.1% | $5.00 | $30.0 | $0.00051 | $0.00306 |
| Claude Opus 4.5 Anthropic | 138 | ≈ est | <0.1% | $5.00 | $25.0 | $0.00069 | $0.00345 |
| Claude Opus 4.6 Anthropic | 185 | ≈ est | <0.1% | $5.00 | $25.0 | $0.00093 | $0.00462 |
| Claude Opus 4.7 Anthropic | 185 | ≈ est | <0.1% | $5.00 | $25.0 | $0.00093 | $0.00462 |
| Claude Opus 4.8 Anthropic | 185 | ≈ est | <0.1% | $5.00 | $25.0 | $0.00093 | $0.00462 |
| Baichuan4 Baichuan | 92 | ≈ cal | 0.3% | $14.1 | $14.1 | $0.00130 | $0.00130 |
| Claude Fable 5 Anthropic | 185 | ≈ est | <0.1% | $10.0 | $50.0 | $0.00185 | $0.00925 |
| Claude Opus 4.1 Anthropic | 138 | ≈ est | <0.1% | $15.0 | $75.0 | $0.00207 | $0.010 |
| GPT-5.4 Pro OpenAI | 102 | exact | <0.1% | $30.0 | $180 | $0.00306 | $0.018 |
| GPT-5.5 Pro OpenAI | 102 | exact | <0.1% | $30.0 | $180 | $0.00306 | $0.018 |
How it works
The counting method
There is no universal token count — each vendor's tokenizer slices text differently, and you're billed by their count. We count with the best method available per vendor and label which one you're seeing: exact, cal (measured calibration), or est (editorial estimate).
/* exact — OpenAI family */
tokens = BPE_encode(text, o200k_base).length
/* cal — measured on the vendor's own published tokenizer (2026-06-10) */
tokens ≈ characters / measured_chars_per_token(vendor)
/* English prose ≈5.1–5.3 c/t · code ≈3.2–3.5 · JSON ≈2.2–2.7 */
/* est — no public tokenizer (Claude, Gemini, Grok, Kimi) */
tokens ≈ characters / editorial_chars_per_token(model)
/* Claude ≤4.6: ~3.5 · Claude 4.7+: ~2.6 · Gemini: ~3.9–4.2 */
cost_as_input = tokens × input_price / 1,000,000
cost_as_output = tokens × output_price / 1,000,000 The cal calibrations were measured on 2026-06-10 by running a fixed English corpus through each vendor's own published tokenizer (downloaded from their official Hugging Face repos: Qwen 3.5, DeepSeek V3.2, GLM-5, MiniMax M2.1, Mistral Nemo, Hunyuan A13B, Seed-OSS, Baichuan M2, Yi 1.5). Closed API tiers measured on an open sibling are an assumption — the closest one available. One measured finding worth knowing: code tokenizes ~35–60% heavier than prose, JSON heavier still.
For the est vendors no public tokenizer exists; estimates follow vendor documentation — Anthropic's own docs state the tokenizer introduced with Opus 4.7 produces roughly a third more tokens than earlier models. GPT-5.x counts use o200k_base — OpenAI hasn't published a newer public encoding. Prices verified 2026-06-09.
Full methodology at /methodology/. Found a count that looks off? Report it.
Quick reference
Words to tokens
English prose averages ≈0.75 words per token (≈4 characters per token). Claude 4.7+ models run roughly a third heavier on identical text.
| Words | Tokens · o200k family | Tokens · Claude 4.7+ |
|---|---|---|
| 100 | ≈ 133 | ≈ 177 |
| 250 | ≈ 333 | ≈ 443 |
| 500 | ≈ 667 | ≈ 887 |
| 1,000 | ≈ 1,333 | ≈ 1,773 |
| 2,500 | ≈ 3,333 | ≈ 4,433 |
Token questions.
What tokens are, why vendors disagree, and where exact counting is (and isn't) possible.
Q · 01 What is a token? +
The unit LLMs read and bill by. A tokenizer splits text into sub-word pieces — common words are often one token, rare words split into several. In English prose one token averages ~4 characters or ~0.75 words, so 1,000 words ≈ 1,330 tokens. Code, non-Latin scripts, and unusual formatting tokenize heavier.
Q · 02 Why do vendors count the same text differently? +
Every vendor trains its own tokenizer vocabulary. The same paragraph can differ by 30%+ between vendors — and even between model generations: Anthropic's tokenizer introduced with Claude Opus 4.7 produces roughly a third more tokens than earlier Claude models for identical text. Since you pay per token, tokenizer choice is a hidden price difference — that's why the counter shows the full per-vendor range.
Q · 03 How accurate are these counts? +
Three tiers, labeled on every row. exact (OpenAI): counted with the o200k_base BPE encoding, the same algorithm the API uses. cal (most vendors): chars-per-token measured on the vendor's own published tokenizer from their official Hugging Face repo — accurate to a few percent on English prose, though heavy code or non-Latin text shifts the ratio. est (Claude, Gemini, Grok, Kimi): no public tokenizer exists, so these follow vendor-documented estimates (±15%). For billing-exact counts on those vendors, use their free count-tokens APIs (Anthropic's
/v1/messages/count_tokens, Gemini's countTokens).Q · 04 Which tokenizer do GPT-5.x models use? +
OpenAI hasn't published an explicit mapping for GPT-5.4/5.5 in
tiktoken yet, but every chat model since GPT-4o uses o200k_base and no newer public encoding exists — so we count GPT-5.x with o200k_base and label it accordingly. If OpenAI ships a new encoding, the counter will switch.Q · 05 Is my text uploaded anywhere? +
No. Counting runs entirely in your browser — the exact OpenAI encoder is downloaded to your browser (a one-time ~1 MB fetch on first keystroke), and your text never leaves the page. No analytics on the text field, no server round-trips. Safe for contracts, code, and anything confidential.
Q · 06 What do the two cost columns mean? +
As input — what it costs to send this text to the model once (your prompt, context, retrieved documents). As output — what it costs if the model generates this text as a response. Output rates run 3–5× input on most models, which is why long generations dominate bills.
Q · 07 Why does the % of context column matter? +
Models can only read a limited window (128K–2M+ tokens depending on the model). The column shows how much of each model's window your text occupies — paste a large document and you'll see which models can actually take it whole, before you spend anything.
Q · 08 What's the quick rule of thumb for words to tokens? +
English prose: tokens ≈ words ÷ 0.75 (or words × 1.33), and 1 token ≈ 4 characters. Add ~33% on top for Claude 4.7+ models, and expect heavier counts for code and non-Latin languages. The reference table above has the common sizes pre-computed.
Embed
Put this counter on your site
Free to embed — counting stays client-side in your readers' browsers too.
More calculators
The cost toolbox
Every tool runs on the same live-pricing backbone. See the full index →