LLM Pricing Comparator
Set input and output tokens per call and your monthly request volume, and instantly compare cost per call and cost per month across GPT-4o, Claude, Gemini and open models — sorted cheapest-first with the spread shown. See why output tokens dominate and how a 10× price gap between models turns into real dollars.
| Model | Input $/1M | Output $/1M | Cost / call | Cost / month | vs cheapest |
|---|---|---|---|---|---|
| GPT-4o mini | $0.15 | $0.6 | $0.000525 | $52.50 | 1.0× |
| Llama 3.1 70B | $0.6 | $0.6 | $0.001200 | $120.00 | 2.3× |
| Gemini 2.5 Flash | $0.3 | $2.5 | $0.001700 | $170.00 | 3.2× |
| Claude Haiku 3.5 | $0.8 | $4 | $0.003200 | $320.00 | 6.1× |
| Gemini 2.5 Pro | $1.25 | $10 | $0.006875 | $687.50 | 13.1× |
| GPT-4o | $2.5 | $10 | $0.008750 | $875.00 | 16.7× |
| Claude Sonnet 4 | $3 | $15 | $0.0120 | $1,200 | 22.9× |
| Claude Opus 4 | $15 | $75 | $0.0600 | $6,000 | 114.3× |
Indicative published list prices, USD per 1M tokens — always confirm against each vendor’s current pricing. Cost / call = input tokens ÷ 1M × input price + output tokens ÷ 1M × output price.
Two prices, not one
Every LLM API quotes two numbers: a price per million input tokens and a price per million output tokens. Input is everything you send — the system prompt, tool definitions, conversation history and retrieved context — and output is only what the model generates. They are billed separately, and output is almost always the pricier of the two, often 3–5× the input rate, because tokens are generated one at a time while input is processed in a single parallel prefill.
The formula
cost per call = input_tokens ÷ 1,000,000 × input_price + output_tokens ÷ 1,000,000 × output_price, then multiply by your monthly request volume. Because prices are per million tokens, a single call looks almost free — until you multiply by millions of requests and the fractions of a cent become a real line item.
Why model choice dwarfs everything else
The gap between a small model and a frontier one is not 20% — it is often 10× to 100×. A task that costs pennies on GPT-4o mini or Claude Haiku can cost dollars on Opus or GPT-4o. The engineering win is rarely shaving tokens; it is routing — sending the easy 90% of calls to a cheap model and reserving the expensive one for the calls that genuinely need it.
How to use this
Set the input and output tokens of a typical call and your real monthly volume, and read the sorted bill. The point isn't the exact dollar figure — list prices drift — but the shape: which models are in your budget, how much the tail of frontier pricing costs, and where a router pays for itself.
How it works
- Cost/call = input tokens ÷ 1M × input price + output tokens ÷ 1M × output price.
- Output tokens usually cost 3–5× more than input — verbosity is expensive.
- Small models can be 10–100× cheaper than frontier ones for the same task.
- Sorted cheapest-first for your exact token mix, not a headline price.
Frequently asked questions
How is LLM API cost calculated?
Cost per call = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price), where prices are quoted per million tokens. Multiply by your monthly request count for the monthly bill. Input includes everything you send — prompt, system message, tools and retrieved context — and output is what the model generates.
Why are output tokens more expensive than input tokens?
Almost every provider charges more per output token — often 3–5× the input price — because generation is sequential and compute-heavy: each output token requires a full forward pass, while input tokens are processed in parallel during the prefill. That is why a chatty, verbose model can cost far more than its input price suggests.
Which LLM is the cheapest?
It depends on your input/output mix, but small models (GPT-4o mini, Claude Haiku, Gemini Flash) are one to two orders of magnitude cheaper than frontier models (Claude Opus, GPT-4o, Gemini Pro). This tool sorts every model by your actual usage so you can see the cheapest one for your workload rather than a headline number.
Are these prices exact?
They are indicative published list prices per million tokens and change often; open models are shown at a typical serverless-host rate. Use the comparison to understand the shape of the trade-off, then confirm the current number against each vendor’s pricing page before committing.