LLM Cost Calculator – Calculate API costs for GPT, Claude & Gemini
Calculate in seconds what your LLM application will cost. Current API prices for OpenAI (GPT-6, GPT-5.x, o3), Anthropic (Claude Fable, Opus, Sonnet, Haiku), Google (Gemini 3 Pro & Flash), Meta (Muse Spark), Mistral, DeepSeek, Amazon Nova and xAI Grok, including prompt caching, monthly estimates and a direct model comparison. Prices are updated live.
What is an LLM cost calculator?
An LLM cost calculator (also called LLM price calculator) works out the API costs of your language model application. Providers like OpenAI, Anthropic and Google bill per token, separately for input tokens (your prompt) and output tokens (the response). Our calculator shows you at a glance whether GPT-6, Claude Sonnet or Gemini Flash is the cheaper choice.
How do I calculate the cost of an LLM API?
The formula:
Cost = (input tokens × input price + output tokens × output price) / 1,000,000
Example Claude Sonnet 5.5 ($2 / $10 per 1M tokens), 1,500 input + 500 output:
- Input: 1,500 × $2 / 1M = $0.003
- Output: 500 × $10 / 1M = $0.005
- Per request: $0.008, at 10,000 requests/month $80
The calculator does this automatically for all models side by side.
Which models are supported?
Prices are loaded live from llm-prices.com, always up to date, currently more than 170 models. Among others:
- OpenAI: GPT-6 (Astra, Sol, Luna), GPT-5.x, GPT-4.1, GPT-4o, o3, o4-mini
- Anthropic: Claude Fable 5.1, Opus 5.5, Sonnet 5.5, Haiku 5.5 and older Claude models
- Google: Gemini 3.x Pro and Flash, Gemini 2.5
- Meta: Muse Spark
- Mistral: Mistral Large, Medium 3, Small 3.1, Magistral, Codestral, Ministral
- DeepSeek: DeepSeek V4 Pro and Flash, DeepSeek Chat and Reasoner
- Amazon: Nova Micro, Lite, Pro, Premier
- xAI: Grok 4.x, Grok 4 Fast, Grok Code Fast
GPT vs. Claude vs. Gemini: which model is the cheapest?
The price differences are huge: a factor of 100 or more between the cheapest and the most expensive model is normal. For orientation (input / output per 1M tokens):
- Budget: Amazon Nova Micro, Ministral 3B, GPT-5 Nano, from $0.035 input
- Mainstream: GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4 Flash, Gemini 3.1 Flash-Lite, $0.10–$0.25 input
- Premium: GPT-6 Sol, Claude Sonnet 5.5, Gemini 3.1 Pro, $2 input, $10–$12 output
- Top tier: Claude Opus 5.5, Claude Fable 5.1, GPT-6 Astra, $4–$10 input, up to $50 output (Pro variants considerably more)
Rule of thumb: the budget tier is enough for classification and tag generation. For code, reasoning and agents you need premium.
What are input, output and cached tokens?
- Input tokens (prompt): everything you send to the model, i.e. system prompt, user message, history, tool definitions
- Output tokens (completion): the response. Typically around 5× more expensive than input
- Cached input tokens: OpenAI, Anthropic and Google offer prompt caching, which makes repeated system prompts up to 90% cheaper
The calculator takes all three into account.
How many tokens does my prompt have?
Before you calculate costs, you need to know how many tokens your text has. For that we built a separate Tokenizer that counts your text exactly for OpenAI models and estimates it for Claude, Gemini & co.
Saving costs: 7 tips for LLM APIs
- Try a smaller model: GPT-6 Luna or Claude Haiku is often enough instead of premium
- Shorten the system prompt: every word costs money on every request
- Enable prompt caching: 50–90% savings on repeated prompts
- Limit the output: set
max_tokens, enforce a JSON schema - Use the Batch API: OpenAI & Anthropic offer 50% off async workloads
- Embeddings instead of LLM calls for retrieval and classification
- Compare prices regularly: new models undercut the old ones almost every month
FAQ
How much does GPT-6 Sol cost per 1,000 tokens?
$0.002 input, $0.01 output (up to 272,000 tokens of context). 1,000 input tokens = 0.2 cents, 1,000 output tokens = 1 cent.
Is Claude cheaper than GPT?
It depends on the model. Claude Sonnet 5.5 and GPT-6 Sol currently cost the same ($2 / $10), as do Claude Haiku 5.5 and GPT-6 Luna ($0.10 / $0.50 with short context). Claude Fable and GPT-6 Astra are the expensive top models. Best to compare them directly in the calculator.
Which LLM is the cheapest?
The budget winners are Amazon Nova Micro, Ministral 3B and GPT-5 Nano. For production use, most teams do best with GPT-6 Luna or Claude Haiku 5.5.
How much does a chatbot cost per month?
A chatbot with 2K input + 500 output on GPT-6 Luna, 10,000 requests: about $4.50/month. On GPT-6 Sol: about $90/month.
How up to date are the prices?
They are fetched live from llm-prices.com; the last update date is shown below the calculator.
Does the calculator cover vision, audio or embeddings?
No, it focuses on text tokens. Image and audio inputs are priced separately.