LLM API Cost Calculator
Price a workload across 149 models from 14 creators.
10M input + 2M output tokens
Enter (1M tokens is about 750K words or 2,500 pages).
Cost by model
For 10M input and 2M output tokens, the cheapest priced model is Llama 3.2 1B at $0.24; the median is $10.40 and the most expensive o1-pro at $2,700.00.
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning. LLM rates are base-tier, on-demand prices per 1M tokens at the cheapest provider listing each model; cached-input, batch and long-context tiers are not included.
What common jobs cost
Each job is a stated assumption about tokens, priced at every model's cheapest provider. Each sentence names the cheapest model we track for that job and the cheapest and most expensive from one of the largest creators, a different creator per job.
Summarising a 200-page book
Summarising a 200-page book (60,000 words in, a two-page summary out: about 80K input tokens and 798 output tokens) costs $0.002 on Llama 3.2 1B, the cheapest model we track; among OpenAI's models it runs from $0.003 on GPT-OSS-20B to $12.45 on o1-pro.
Translating a 5,000-word article
Translating a 5,000-word article (the text in and a translation of the same length out: about 7K input tokens and 7K output tokens) costs under $0.001 on Llama 3.2 1B, the cheapest model we track; among Alibaba Cloud's models it runs from $0.002 on Qwen3-Coder-30B-A3B to $0.058 on Qwen3.8-2.4T-A95B.
Classifying 100,000 support tickets
Classifying 100,000 support tickets (150 words each, a one-word label back: about 20M input tokens and 100K output tokens) costs $0.40 on Llama 3.2 1B, the cheapest model we track; among Google Cloud's models it runs from $1.01 on Gemma 3 12B Instruct to $41.20 on Gemini 3.1 Pro Preview.
Running a chatbot through 10,000 conversations
Running a chatbot through 10,000 conversations (five turns of 200 words each way, with the history resent every turn: about 66M input tokens and 13M output tokens) costs $1.60 on Llama 3.2 1B, the cheapest model we track; among DeepSeek's models it runs from $2.39 on DeepSeek OCR-2 to $382.38 on DeepSeek R1.
Extracting fields from 1,000 invoices
Extracting fields from 1,000 invoices (two pages each, 200 tokens of JSON back: about 798K input tokens and 200K output tokens) costs $0.020 on Llama 3.2 1B, the cheapest model we track; among OpenAI's models it runs from $0.052 on GPT-OSS-20B to $239.70 on o1-pro.
Summarising 50 hours of meeting transcripts
Summarising 50 hours of meeting transcripts (150 spoken words a minute in, a one-page summary per hour out: about 598K input tokens and 20K output tokens) costs $0.012 on Llama 3.2 1B, the cheapest model we track; among Alibaba Cloud's models it runs from $0.041 on Qwen3-Coder-30B-A3B to $1.62 on Qwen3.8-2.4T-A95B.
Reviewing a 10,000-line pull request
Reviewing a 10,000-line pull request (8 tokens a line in, a 1,000-word review out: about 80K input tokens and 1K output tokens) costs $0.002 on Llama 3.2 1B, the cheapest model we track; among Google Cloud's models it runs from $0.004 on Gemma 3 12B Instruct to $0.18 on Gemini 3.1 Pro Preview.
Common questions
How is the cost calculated?
Input tokens times the model's input rate, plus output tokens times its output rate, both per 1M tokens. The rate is the cheapest provider listing the model, with input and output taken from that one provider. Cached-input discounts, batch pricing, long-context surcharges and reasoning-token billing are not modelled, because the rates we track do not include them; a workload that uses them will cost less, or more, than the figure here.
How many words is 1 million tokens?
About 751,880 English words, or roughly 2,506 book pages at 300 words a page. This page converts at 1.33 tokens per word, OpenAI's rule of thumb that a token is about three quarters of a word; Gemini quotes 60 to 80 words per 100 tokens. Claude 4.7 and later tokenize about 30% denser than earlier Claude models, and text in languages other than English can take 1.5x to 3x the tokens, so treat the conversion as an estimate and count with the provider's tokenizer before relying on it.
How many tokens are in a 200-page book?
About 80K tokens: 200 pages at 300 words is 60,000 words, times 1.33 tokens a word. A 300-page book is about 120K tokens and a typical page about 399. Dense technical text, tables and code run higher.
What does 1 million tokens cost?
Per 1M input tokens, the models on this page run from $0.020 (Llama 3.1 8B) to $150.00 (o1-pro); the median is $0.60. Output tokens cost more, typically 3x to 8x the input rate, so a job that writes as much as it reads costs several times a job that only reads.
Which provider's price is used for each model?
The cheapest one listing it. Open-weight models are served by several providers at different rates, sometimes severalfold apart; closed models resold through partners are usually at the creator's list price. Every model name links to its page, which lists every provider we track for it.
How current are these prices?
Rates were last updated on August 26, 2026 and are reviewed against each provider's published pricing page by hand. The full table, with every provider, is on the LLM API pricing page.