LLM API pricing comparison and token cost calculator

Your workload

Reset

Start from a preset that looks like your case, then adjust the numbers. A token is about ¾ of an English word: 1,000 tokens ≈ 750 words.

Presets

Website chatbot Short questions, short answers, many visitors 500 in · 200 out · 50k/mo · 50% cache Support assistant Ticket plus some history in, a reply out 2.5k in · 500 out · 10k/mo · 40% cache Email & copy drafting Brief in, a few paragraphs out 800 in · 600 out · 3k/mo Translation Same length in and out 1.5k in · 1.5k out · 5k/mo Document summaries A 15-page document in, one page out 12k in · 800 out · 2k/mo RAG Q&A over your docs Question plus retrieved passages, short answer 6k in · 400 out · 20k/mo · 60% cache Data extraction Form or invoice in, structured JSON out 3k in · 300 out · 30k/mo · 30% cache Coding assistant Large code context per turn, like Cursor or Claude Code 25k in · 2.5k out · 5k/mo · 80% cache Long-document analysis Whole contracts or reports, a few times a day 100k in · 1k out · 500/mo

Show all 9 presets

Measure usage in

TokensWordsCharacters

Input tokens per requestWhat you send: prompt, documents, chat history. Output tokens per requestWhat the model writes back. Usually 3–5× pricier than input. Requests per month

Usage period

Per monthPer day

Cached input:

40

%Share of input tokens served from the provider's prompt cache (repeated system prompts, documents, history). Applied only where a vendor publishes a cached price. Copy shareable link

Your workload, filters and shortlist are stored in the URL.

Cost meter

—

Input tokens—

Output tokens—

Requests / mo—

Cached input—

All vendors Anthropic DeepSeek Google Meta Llama OOpenAI xAI

Sort Lowest costHighest costInput priceOutput priceName Cache pricing only

TableCost chart

Shortlist	Model	Input / 1M	Output / 1M	Cached / 1M	Context	Per request	Per month	Source

Not included yet

• OpenAI — 403 on openai.com (not retried); platform.openai.com/docs/pricing was readable and used (official page)
• Mistral — Per-model API prices are JS-rendered; only one FAQ example price in HTML (official page)
• Cohere — Current Command model token prices not in HTML; only legacy models listed in FAQ (skipped) (official page)
• Mistral — per-model prices load via JavaScript and are not in the page HTML (official page)

How the numbers are calculated

• Per request = input tokens × input price + output tokens × output price. The cached share of input is billed at the vendor's cached-input price when one is published; otherwise at the normal input price.
• Per month = per request × requests per month (per day × 30 when you switch to daily). Cache write fees, batch discounts, long-context surcharges, rate-limit tiers and taxes are not included; follow the source link for the vendor's full terms.
• Words are converted at 1 word ≈ 1.33 tokens and characters at 4 characters ≈ 1 token; real tokenizers vary by language and content.
• Prices are copied from official vendor pages on the date shown and may change. Rows without a published cached price show “—”.