TokenCalculator.dev is a privacy-first, browser-based LLM token counting and API cost estimation tool that helps developers and teams understand the token size and cost of an input payload before making an API call.
Purpose
- Measure the number of input tokens for text, code, structured data, readable documents, and supported images across multi-provider models.
- Estimate input cost and check whether a prompt exceeds a model's context window.
- Compare costs across different providers and models under real workloads, rather than comparing models alone.
Key capabilities
- Single input, cross-provider measurement: after pasting a prompt or attaching files, you can simultaneously view token results, context fit, and estimated input cost for 43 current models (from 9 providers: OpenAI, Anthropic, Google, DeepSeek, Qwen, Moonshot Kimi, xAI Grok, Z.ai GLM, Meta).
- File support: supports text-based PDF, DOCX, as well as plain text, Markdown, CSV, JSON, JavaScript, TypeScript, Python, Go, Rust, Java, C/C++, HTML, CSS, XML, YAML, TOML, SQL, and shell files, up to 12 MB per file. Scanned PDFs require OCR and are explicitly rejected, rather than returning misleading zero-token results.
- Image token estimation: supports PNG, JPEG, WebP, GIF, reads width and height locally, and calculates vision tokens according to the published tile, patch, or media-token rules of the selected OpenAI, Claude, and Gemini models. Image tokens are added to document text and pasted prompt text, making it easier to plan multimodal requests. If a model has no published vision rules, the result shows "image unavailable" rather than fabricating an estimate.
- Visible calculation provenance: each result indicates the method used and the accuracy level, distinguishing exact BPE tokenization from deterministic provider-calibrated projection.
- Pricing data: from Models.dev, with a bundled offline fallback registry; prices are listed per 1 million tokens for input and output separately, and cached input is the discounted price providers charge for reused prompt prefixes.
- What is reported: input token count, context window fit, and estimated input cost.
How to use
- Paste the prompt into the webpage, or attach PDF, DOCX, code, data files, and images.
- Select a provider card to view each available model.
- Results and costs update in real time with the input, with no need to upload content.
- Extraction for text, code, PDF, and DOCX, image dimension reading, and token calculation all run locally in the browser; except for public model metadata that may be refreshed from Models.dev, no prompt, document content, or uploaded image is sent to TokenCalculator.dev or any calculation server.
Use cases
- Price knowledge base ingestion tasks: extract text-based PDFs locally, measure the full payload, and compare context fit and input cost across OpenAI, Claude, Gemini, and DeepSeek.
- Budget requests for vision models: combine image dimensions with prompt text and apply the published tile, patch, or media-token rules of each vision-capable model.
- Measure code repository review load: attach code, configuration, and documentation files as a single workload, then choose a model with sufficient context capacity.
- Compare batch economics: measure structured data together with instructions, and compare input rates across providers before committing to a recurring workload.
Limitations
- Scanned or encrypted PDFs require OCR, which this tool does not support and will explicitly reject.
- Image counting is an image token and cost estimate, not OCR; pixel data is not uploaded, retained, or used to extract text from images.
- Providers may add message wrapping, system instructions, tools, or model-specific processing, so the final bill may be affected; provider documentation remains the authoritative source. Long-context tiers, batch discounts, and provider-side overhead may change the actual invoice amount.
- Non-OpenAI text results are explicitly labeled as estimates unless the provider publishes a deterministic local formula.
- The current material does not provide more information about billing models, pricing plans, or registration requirements.
