Developer Tools
AI Token Cost Calculator (Claude, GPT, Gemini)
Estimate the cost of an LLM workload. Plug in tokens-in, tokens-out, monthly volume, and get a per-call and monthly cost across major providers.
%
On input tokens
Per call
$0.0180
In: $0.00600 · Out: $0.01200
Monthly
$180
10,000 calls
Annualized
$2,160
Cost comparison
All models · same workload · sorted cheapest first
Monthly cost @ 10,000 calls
Llama 4 Maverick (hosted)
Meta
$12
DeepSeek-V3
DeepSeek
$14
Gemini 2.5 Flash
Google
$26
Claude Haiku 4.5
Anthropic
$60
GPT-5 mini
OpenAI
$65
GPT-4o
OpenAI
$130
Gemini 2.5 Pro
Google
$170
Claude Sonnet 4.6
Anthropic
$180
GPT-5
OpenAI
$780
Claude Opus 4.7
Anthropic
$900
How this is calculated
For every model, we calculate:
cost = (input_tokens × input_price + output_tokens × output_price) / 1,000,000
If you set a cache hit ratio, that fraction of input tokens is billed at the lower cache-read rate (Anthropic, OpenAI, and Google all expose one).
Token estimation rule of thumb
- ~4 characters per token for English
- ~0.75 tokens per word
- 1,000 words ≈ 1,300–1,500 tokens (input)
- Code is heavier — closer to 3 chars/token
Why caching matters
System prompts, tool definitions, and long context can be cached at 10–25% of the read price. For chat workloads with stable system prompts, caching cuts bills 60–80%.
Frequently asked
- Prices reflect public list pricing as of the page's last-updated date. Always confirm with your provider — usage tiers and discounts vary.
- Yes for Anthropic and OpenAI. Cached input tokens are billed at the lower cache-read rate when applicable.