Calculator › Sources & method

Sources and methodology

This site exists to answer one question quickly and honestly: what will this workload cost on each model? Here is exactly how the numbers are produced.

Price sources

All prices were read by hand from each provider's official pricing page on October 19, 2026:

• OpenAI: https://platform.openai.com/docs/pricing. Standard tier. Prompts over 272K input tokens are billed at the long-context rate shown. Batch API: 50% off. Regional (data-residency) endpoints add 10% on newer models.
• Anthropic: https://docs.anthropic.com/en/docs/about-claude/pricing. Cached input = cache-hit price. Writing to the cache costs 1.25x input (5-minute) and is not included. Claude 4.7 and later use a newer tokenizer that produces about 30% more tokens for the same text. Batch API: 50% off.
• Google: https://ai.google.dev/gemini-api/docs/pricing. Paid tier, text input. Prompts over 200K tokens are billed at the higher rate shown (Pro models). Batch Mode: 50% off. Google's page was last updated 2026-10-16.
• xAI: https://docs.x.ai/docs/models. Prompts of 200K tokens or more are billed at the higher rate shown, for every token in the request.
• DeepSeek: https://api-docs.deepseek.com/quick_start/pricing. Standard (peak) prices. Off-peak hours are 50% off.

Prices are standard pay-as-you-go list prices in USD for text. Regional or data-residency endpoints (usually +10%), priority or "fast" modes, enterprise discounts, free tiers, and cloud-marketplace prices (Azure, Bedrock, Vertex AI) are not included.

How costs are calculated

• Per request: uncached input × input price + cached input × cached price + output × output price, divided by 1,000,000.
• Long-context pricing: when a prompt exceeds the model's threshold, the higher rate applies to the whole request.
• Batch: 50% off input, output and cached input where the provider publishes a batch discount (OpenAI, Anthropic, Google). For some older OpenAI models the batch rate has no separate cached price, so cached input is billed at the batch input rate.
• DeepSeek off-peak: 50% off everything.
• Tokenizer adjustment: Claude 4.7 and later models are charged for 1.3× the tokens, per Anthropic's note that their newer tokenizer produces about 30% more tokens for the same text.
• A month is 30 days.

Not included

• Prompt cache write charges (Anthropic: 1.25× input for 5-minute writes, 2× for 1-hour; OpenAI publishes cache-write prices for its newest models), and Google's cache storage fee.
• Tool fees such as web search, code execution containers or grounding.
• Image, audio and video tokens, embeddings and fine-tuning.
• Taxes.

Token counting

The counter uses OpenAI's open-source o200k_base tokenizer (via the js-tiktoken library), loaded on demand and run entirely in your browser. Other providers' tokenizers are not public in the same way, so their counts are estimates.

Price change log

• October 19, 2026: all 47 models checked. Upcoming changes flagged: GPT-5.6 Sol promotional price (through Nov 21, 2026), Gemini 4 Argon (doubles Dec 1, 2026), Gemini 3.8 Flash (doubles Jan 1, 2027).