OneTaskBy Tim King · Updated
LLM API Pricing Comparison
Input, Cached and Output Prices Per Million Tokens for 82 Models
OneTask is free, with no sign-up, and built by one person.
If this tool saved you time, a coffee helps keep new ones coming.
How the LLM API Prices works
This LLM API pricing comparison lists the standard price of 82 language models from Anthropic, OpenAI, Google, Mistral, DeepSeek and xAI, all in US dollars per million tokens. Search for a model or a vendor to see its input, cached input and output price side by side, with a link to the pricing page each figure came from.
How to read an LLM API pricing comparison
Every vendor bills text models by the token, and quotes the price per million tokens (often written MTok). Input tokens are everything you send: the system prompt, the conversation so far and any documents. Output tokens are what the model writes back, and they cost several times more. A request with 20,000 tokens in and 1,000 out on a $3 / $15 model costs 20,000 × $3 ÷ 1,000,000 plus 1,000 × $15 ÷ 1,000,000, which is $0.06 plus $0.015, or $0.075.
Cached input
When the start of a prompt repeats between calls, most vendors can reuse it and bill it at a much lower cached rate, often a tenth of the normal input price. That matters for chat apps and agents, where the same long history is resent on every turn. Some vendors also charge to write the cache; Anthropic, for example, bills a five-minute cache write at 1.25 times the input price. A dash means the vendor lists no cached price for that model.
What the table leaves out
These are standard pay-as-you-go prices. Batch APIs usually halve them, priority and fast modes raise them, and some vendors charge more for prompts over 200,000 tokens, which the Notes column records. Tokens are not the same size across vendors either: the same paragraph can be 30% more tokens on one tokenizer than another, so compare real bills rather than list prices when the gap is small. Image, audio, embedding and speech models are not included.
How current the prices are
Every price was read from the vendor's own page on the retrieval date shown below the table. Vendors change prices often and without much notice, so check the linked source before committing to a budget.
FAQ
What is the cheapest LLM API per million tokens?
In this table the lowest list prices belong to small models such as gpt-5-nano, Gemini 2.5 Flash-Lite and Ministral 3 3B, all around ten cents or less per million input tokens. They suit classification, extraction and short replies, but larger models are usually worth the cost for reasoning or long writing.
Why do output tokens cost more than input tokens?
Input tokens are processed in one parallel pass, while output tokens are generated one at a time, each needing a full pass through the model. That sequential work uses far more compute per token, so vendors price output at roughly four to eight times the input rate.
How do I estimate the cost of a prompt?
Count the tokens, which is roughly four characters or three quarters of a word in English, then multiply input tokens by the input price and output tokens by the output price, and divide each by one million. Add them together, and use the cached rate for any repeated prefix.
See also: all Tech Data tools, Everyday Data.
Related tools
Word Counter
Words, six readability scores, keyword density and platform limits.
Character Counter
Characters against SMS, meta and social post limits.
JSON Formatter
Format, minify and validate JSON, with located errors.
Currency Converter
Convert between 30 currencies at the mid-market rate.
LLM Cost Calculator
Price LLM API calls per request and per month, with caching.