Vibe Engines is free, interactive AI engineering & system design — runs in your browser. Pick a learning path
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 — last reviewed 2026-07; always confirm against each vendor’s current pricing. Cost / call = input tokens ÷ 1M × input price + output tokens ÷ 1M × output price.
Every LLM API quotes two prices, one per million input tokens and one per million output tokens, and the mix of the two decides which model is cheap for you. This comparator applies both prices to a typical call of yours, multiplies by your monthly volume and sorts eight models from cheapest to most expensive.
What it computes
cost per call = input tokens / 1,000,000 × input price
+ output tokens / 1,000,000 × output price
monthly = cost per call × requests per month
vs cheapest = monthly / cheapest model’s monthly
spread = most expensive monthly / cheapest monthly- Input tokens are everything sent with the call: system prompt, tool definitions, history and retrieved context.
- Output tokens are what the model writes back.
- Prices are the tool’s built-in illustrative list prices in US dollars per million tokens, last reviewed in July 2026.
A worked example
The defaults are 1,500 input tokens and 500 output tokens per call, at 100,000 requests a month. For GPT-4o mini, at $0.15 in and $0.60 out: 0.0015 × 0.15 + 0.0005 × 0.60 = $0.000525 per call, or $52.50 a month. For Claude Opus 4, at $15 and $75: 0.0015 × 15 + 0.0005 × 75 = $0.06 per call, or $6,000 a month.
| Model | In / out per 1M | Per call | Per 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× |
The spread is 114×. Output is only a quarter of the tokens here but most of the cost on every proprietary model: on Opus, $0.0375 of the $0.06. The ranking also depends on the mix. Raise input to 20,000 tokens, typical of a retrieval-heavy prompt, and Gemini 2.5 Flash moves ahead of the hosted Llama, because its input price is half as high.
Assumptions and limits
- List prices only. Prompt caching, batch discounts, committed-use contracts and free tiers can cut real bills substantially and are not applied.
- One rate per model. Some providers charge more above a prompt-length threshold, and reasoning models bill their hidden thinking tokens as output.
- Token counts differ by model. Each family has its own tokenizer, so the same text can be a different number of tokens for each.
- Price is not quality. The cheapest model may fail your task; the tool does not measure accuracy, latency or rate limits.
- Prices age. Models are retired and repriced often, so treat the table as a dated snapshot.
How to use the result
- Measure a typical call from real logs, including the system prompt and context, rather than guessing.
- Shortlist by quality, then price. Run your evaluation set on two or three candidates and pick the cheapest that passes.
- Route. Sending easy requests to a small model and only hard ones to a frontier model often captures most of the savings with little quality loss.
- Cap output. Shorter answers and a sensible output limit cut the most expensive half of the bill.
How it works
- Cost/call = input tokens ÷ 1M × input price + output tokens ÷ 1M × output price.
- Output tokens usually cost several times more than input; verbosity is expensive.
- Small models can be 10–100× cheaper than frontier ones for the same call.
- Sorted cheapest-first for your exact token mix; prices are illustrative defaults.
Frequently asked questions
How is LLM API cost calculated?
Cost per call is the input tokens divided by one million times the input price, plus the output tokens divided by one million times the output price, because providers quote prices per million tokens. Multiply by your monthly request count for the monthly bill. Input includes everything you send, such as the prompt, system message, tool definitions and retrieved context, and output is only what the model generates in reply.
Why are output tokens more expensive than input tokens?
Generation is sequential: each output token needs its own pass through the model, while all the input tokens are processed together in one parallel prefill step that uses the hardware far more efficiently. Providers price that difference in. In this tool’s illustrative table, output costs four to eight times the input price for the proprietary models and the same as input for the hosted open model, so a verbose answer can dominate the bill.
Which LLM is the cheapest?
It depends on your input and output mix, so the tool sorts every model by your actual usage rather than by a headline price. With the default 1,500 input and 500 output tokens per call, GPT-4o mini is cheapest at about $0.0005 a call, and Claude Opus 4 costs about 114 times as much. Small models are commonly one to two orders of magnitude cheaper than frontier models, which is why routing easy requests to them pays off.
How much does 100,000 LLM API calls cost?
At the default 1,500 input and 500 output tokens per call and the tool’s illustrative prices, 100,000 calls a month range from about $52.50 on GPT-4o mini and $120 on hosted Llama 3.1 70B, through $875 on GPT-4o and $1,200 on Claude Sonnet 4, to about $6,000 on Claude Opus 4. Change the token counts and the order can shift, because each model weighs input and output differently. Confirm current prices before budgeting.
Are these prices exact?
No. They are illustrative list prices per million tokens, last reviewed in July 2026, and vendors change prices and release new models often; the open model is shown at a typical serverless-host rate. The tool also ignores discounts such as prompt caching and batch processing, and tiered rates for very long prompts. Use the comparison to understand the shape of the trade-off, then confirm the numbers against each vendor’s pricing page before committing.