Token counter for GPT, Claude and Gemini
Paste any text or a file and see it split the way a model reads it.
What this text costs and how much room it takes
Input price for sending this text once, and the share of each model's context window it fills.
The count uses OpenAI's published o200k_base tokenizer, the one GPT-4o, GPT-4.1, GPT-5 and the o-series use, running in your browser. OpenAI has not published the tokenizer of its GPT-6 models, and Claude, Gemini and the others use their own, so the per-model figures are estimates (marked ≈). For an exact Claude or Gemini count, Anthropic and Google offer token-counting endpoints in their APIs. Nothing you paste or open is uploaded.
Prices and limits checked on October 26, 2026 on each provider's own page. See the numbers and sources
The numbers behind this tool
Read on October 26, 2026 from each provider's own page. Prices are US dollars per million tokens, standard tier.
| Model | API id | Input | Cached input | Output | Context window | Max output | Source |
|---|---|---|---|---|---|---|---|
| GPT-6 Astra OpenAI | gpt-6-astra |
$10.00 | $1.00 | $50.00 | 1,050,000 | 128,000 | developers.openai.com |
| Flagship ('Our most capable model'). Prices for <=272K input tokens; above 272K: $20.00 in / $2.00 cached / $75.00 out. Context stated as '1.05M', max output as '128K tokens'. Context/max output from developers.openai.com/api/docs/models. | |||||||
| GPT-6.1 Sol OpenAI | gpt-6.1-sol |
$2.00 | $0.10 | $10.00 | 1,050,000 | 128,000 | developers.openai.com |
| Mid option ('Near-Astra performance for complex work at a lower cost'). Above 272K input: $4.00 / $0.20 / $15.00. Context stated as '1.05M', max output '128K tokens'. | |||||||
| GPT-6 Luna OpenAI | gpt-6-luna |
$0.10 | $0.01 | $0.50 | 1,050,000 | 128,000 | developers.openai.com |
| Small ('Our most efficient model for focused, high-volume tasks'). Above 272K input: $0.20 / $0.02 / $0.75. Context stated as '1.05M', max output '128K tokens'. | |||||||
| Claude Fable 5.1 Anthropic | claude-fable-5-1 |
$10.00 | $0.25 | $50.00 | 1,000,000 | 128,000 | platform.claude.com |
| Top model ('demanding reasoning and long-horizon agentic work'). Cached = cache hits/refreshes (0.025x input); 5m cache write $12.50, 1h write $20. Full 1M context at standard pricing. Context stated as '1M tokens', max output '128K tokens'. Claude Mythos 5.1 has the same prices but limited availability. Context/ID from platform.claude.com/docs/en/about-claude/models/overview. | |||||||
| Claude Opus 5.5 Anthropic | claude-opus-5-5 |
$4.00 | $0.20 | $20.00 | 1,000,000 | 128,000 | platform.claude.com |
| Anthropic's recommended default ('start with Claude Opus 5.5 for most workloads'). Cache hits 0.05x input; 5m write $5, 1h write $8. Full 1M context at standard pricing. | |||||||
| Claude Sonnet 5.5 Anthropic | claude-sonnet-5-5 |
$2.00 | $0.10 | $10.00 | 1,000,000 | 128,000 | platform.claude.com |
| Mid ('best combination of speed and intelligence'). Cache hits 0.05x input; 5m write $2.50, 1h write $4. Full 1M context at standard pricing. | |||||||
| Claude Haiku 5.5 Anthropic | claude-haiku-5-5 |
$0.10 | $0.01 | $0.50 | 1,000,000 | 128,000 | platform.claude.com |
| Small. Prices are for prompts up to 100,000 tokens; prompts over 100,000 tokens: $0.50 in / $0.05 cached / $2.50 out (5m write $0.625, 1h write $1). | |||||||
| Gemini 4 Argon | gemini-4-argon |
$2.00 | $0.10 | $10.00 | 1,048,576 | 65,536 | ai.google.dev |
| Flagship ('Our flagship reasoning model'). Paid tier only (not on free tier). Page announces new prices from Dec 1, 2026: $4.00 in / $0.20 caching / $20.00 out. No >200k tier listed. Context window is a unified input+output limit of 1,048,576; default sync output 65,536, up to 1,048,576 with long decode. Context caching also has a storage charge. Limits from ai.google.dev/gemini-api/docs/models/gemini-4-argon. | |||||||
| Gemini 3.8 Flash | gemini-3.8-flash |
$0.75 | $0.075 | $3.75 | 1,048,576 | 65,536 | ai.google.dev |
| Mid. Free tier available. Page announces new prices from Jan 1, 2027: $1.50 in / $0.15 caching / $7.50 out. Context caching also has a storage charge. Limits from ai.google.dev/gemini-api/docs/models/gemini-3.8-flash. | |||||||
| Gemini 3.5 Flash-Lite | gemini-3.5-flash-lite |
$0.30 | $0.03 | $2.50 | 1,048,576 | 65,536 | ai.google.dev |
| Small. Free tier available. Note: older gemini-2.5-flash-lite ($0.10/$0.40) and gemini-3.1-flash-lite ($0.25/$1.50) are cheaper and still listed. Limits from ai.google.dev/gemini-api/docs/models/gemini-3.5-flash-lite. | |||||||
| DeepSeek-V4-Pro (DeepSeek-V4-Pro-0813) DeepSeek | deepseek-v4-pro |
$1.32 | $0.044 | $3.96 | 1,000,000 | 384,000 | api-docs.deepseek.com |
| Flagship. Prices are PEAK rates (cache-miss input). Off-peak rates are half: $0.66 in / $0.022 cached / $1.98 out. Peak = 01:00-04:00 and 06:00-10:00 UTC Mon-Fri excluding Chinese public holidays. Context stated as '1M', max output 'Up to 384K'. Only two models are listed, so there is no separate mid tier. | |||||||
| DeepSeek-V4.1-Flash (DeepSeek-V4.1-Flash-1017) DeepSeek | deepseek-flash |
$0.30 | $0.006 | $1.20 | 1,000,000 | 384,000 | api-docs.deepseek.com |
| Small/fast. PEAK rates; off-peak half: $0.15 / $0.003 / $0.60. Legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are routed here. Context stated as '1M', max output 'Up to 384K'. | |||||||
| Grok 4.7 xAI | grok-4.7 |
$2.00 | $0.50 | $6.00 | 500,000 | – | docs.x.ai |
| Newest Grok listed. Prices for <200k prompt tokens; at >=200k prompt tokens the whole request is billed $4.00 / $1.00 / $12.00. Context stated as '500k'. Max output not stated. The page does not label flagship/mid/small tiers. | |||||||
| Grok 4.3 xAI | grok-4.3 |
$1.25 | $0.20 | $2.50 | 1,000,000 | – | docs.x.ai |
| Cheaper Grok option (grok-4.20-0309 reasoning/non-reasoning have the same prices). >=200k prompt tokens: $2.50 / $0.40 / $5.00. Context stated as '1M'. Max output not stated. Tier label (mid) is our choice, not xAI's. | |||||||
| Mistral Large 4 Mistral | mistral-large-4-0 |
$1.36 | $0.14 | $4.18 | 1,000,000 | – | docs.mistral.ai |
| Flagship (open-weight, v26.10). ID inferred from the docs URL; page displays 'mistral-large-4+1'. Context stated as '1M'. CONFLICT: mistral.ai/pricing states 'Mistral Large costs $0.5 /M tokens in and $1.5 /M tokens out' (version not named). Docs price used here; recheck before publishing. | |||||||
| Mistral Medium 3.5 Mistral | mistral-medium-3-5-26-04 |
$1.50 | – | $7.50 | 256,000 | – | docs.mistral.ai |
| Mid (v26.04). ID inferred from the docs URL; page displays 'mistral-medium-3-5+2'. Verify exact API ID. Context stated as '256k'. Input/output labels taken from docs.mistral.ai/inference/pricing; model page shows the two numbers unlabeled. Cached price not stated. | |||||||
| Mistral Small 4 Mistral | mistral-small-2603 |
$0.15 | – | $0.60 | 256,000 | – | docs.mistral.ai |
| Small (v26.03, Apache 2.0). Page displays ID 'mistral-small-2603' with an unexplained '+1'. Context stated as '256k'. Input/output labels from docs.mistral.ai/inference/pricing. Cached price not stated. | |||||||
- Words per token: "A token is about 4 characters; 100 tokens are about 60-80 English words." (ai.google.dev)
- Tokenizer: o200k_base, gpt-tokenizer 3.4.0 (MIT). Used by GPT-4o, GPT-4.1, GPT-5 and o-series per tiktoken model.py. tiktoken 0.14.0 does not map GPT-6 models to an encoding, so counts for them are estimates. (github.com)
How it counts
The counter runs OpenAI’s published o200k_base tokenizer in your browser. That is the tokenizer GPT-4o, GPT-4.1, GPT-5 and the o-series models use, so for them the count is exact. Each coloured chip in the split view is one token, exactly as that tokenizer cuts the text.
OpenAI has not published the tokenizer of its GPT-6 models, and Anthropic, Google, DeepSeek, xAI and Mistral each use their own. The per-model list therefore marks their figures with ≈: they are the o200k count priced at that model’s rate, a close guide rather than the provider’s own number.
Questions
Why do code and other languages use more tokens?
Tokenizers are built from a large mix of text, and common English words often become a single token. Rare words, code symbols and many non-English scripts break into more pieces. The characters-per-token tile shows this for your own text: the lower it is, the more tokens the text needs.
Does the cost include the answer?
No. The list prices the text you send, once, at each model’s input rate. Answers are billed at the output rate, which is higher on every model listed. The API cost calculator prices both.
Can I count a PDF or a Word file?
The counter reads plain-text files. Copy the text out of a PDF or Word document and paste it, or save it as .txt first.
More tools
How long is a 4,000-token answer?
Words to tokens
ConvertWhat will 10,000 chats a month cost?
API cost calculator
Estimate my billIs a monthly AI plan cheaper than paying per use?
Subscription or API?
Compare for my usageWill my whole document fit?
Context window checker
Check a documentCan my GPU run this model?