Pricing Calculator

Compare costs across the latest LLM providers and models — Claude, GPT-5, Gemini 3, Grok, Llama 4 — to optimize your budget. Prices per 1M tokens, updated June 2026.

Usage Parameters

Total Input Tokens
1,000,000
Total Output Tokens
500,000
Total Tokens
1,500,000
Requests
1,000
RankModelProviderInput PriceOutput PriceTotal Cost
#1
GPT-5 nanoBest Value
OpenAI$0.05/1M$0.4/1M
$0.2500
#2
Llama 4 Scout
Meta$0.11/1M$0.34/1M
$0.2800
#3
Gemini 3 Flash-Lite
Google$0.1/1M$0.4/1M
$0.3000
#4
Grok 4 Fast
xAI$0.2/1M$0.5/1M
$0.4500
#5
DeepSeek V3.2
DeepSeek$0.28/1M$0.42/1M
$0.4900
#6
Llama 4 Maverick
Meta$0.35/1M$1.15/1M
$0.9250
#7
GPT-5 mini
OpenAI$0.25/1M$2/1M
$1.2500
#8
Gemini 3 Flash
Google$0.5/1M$3/1M
$2.0000
#9
Claude Haiku 4.5
Anthropic$1/1M$5/1M
$3.5000
#10
Mistral Large 3
Mistral$2/1M$6/1M
$5.0000
#11
GPT-5.1
OpenAI$1.25/1M$10/1M
$6.2500
#12
Gemini 3 Pro
Google$2/1M$12/1M
$8.0000
#13
Claude Sonnet 4.6
Anthropic$3/1M$15/1M
$10.5000
#14
Grok 4
xAI$3/1M$15/1M
$10.5000
#15
Claude Opus 4.8
Anthropic$5/1M$25/1M
$17.5000
#16
GPT-5.5
OpenAI$5/1M$30/1M
$20.0000
Prices are approximate and may change. Always check the official provider documentation for current pricing.

How LLM API pricing works

Almost every LLM API uses the same billing model: you pay per million tokens, with separate rates for input (the prompt you send) and output (the text the model generates). The gap between providers is large — a frontier model can cost twenty to a hundred times more per token than a small, fast model — so the right choice depends heavily on your traffic and how much of it genuinely needs top-tier reasoning.

This calculator multiplies your input and output token counts by each model's published rate and ranks the results, so you can see the cheapest option for a given workload at a glance. Because most production workloads are dominated by either input (long prompts, RAG context) or output (long generations), the cost ordering can flip depending on your ratio — which is exactly what the calculator is built to reveal.

How to use this tool

  1. Enter the average input tokens, output tokens, and number of requests for your workload.
  2. Review the ranked table — the cheapest model for your specific input/output mix is highlighted as best value.
  3. Compare a frontier model against a smaller one to see how much you'd save by routing simpler tasks to a cheaper tier.

Frequently asked questions

Why is one model cheapest for my workload but not another?+

Because input and output are priced differently. A model with cheap input but pricey output wins on long-prompt, short-answer tasks and loses on short-prompt, long-answer tasks. Always calculate with your real token ratio.

Can I cut costs without changing models?+

Yes. Prompt caching, the Batch API (typically 50% off), shorter system prompts, and trimming output length all reduce spend. Routing easy requests to a smaller model is usually the biggest lever.

Are these prices current?+

We review them regularly, but model pricing changes often. Treat the figures as estimates and confirm the live rate on the provider's pricing page before committing a budget.