LLM API Price Comparison

Enter your monthly token usage and see every major LLM API ranked cheapest first — GPT, Claude, Gemini, DeepSeek — with your real, blended cost per 1M tokens. The list changes with your input/output mix, so you get the cheapest model for your workload, not a generic chart.

🔒 100% in-browser — nothing you enter is uploaded

Your monthly usage

Avg prompt size per request.
Avg reply length per request.
Total calls across your app.
Cheapest for you
$0
—
Priciest listed
$0
—
Model $/1M in $/1M out Eff. $/1M
(your mix)		 Your $/mo

Prices are per 1M tokens, standard (non-cached, non-batch) tier, verified June 2026 — always confirm live rates at OpenAI, Anthropic, Google, and DeepSeek. Effective $/1M blends input and output at your ratio, so it reflects what you'd actually pay per million tokens, not the headline input price. The Local (ABUZ8 OS) row is $0 per token — marginal cost is electricity only; the fixed hardware cost is covered in the self-host vs cloud calculator.

The cheapest token is the one you don't rent.

Every row above is a bill that scales with your usage. ABUZ8 OS runs open models on hardware you own for $0 per token, and routes to a paid API only for the few calls a local model can't clear — sovereign by default, cloud only when it has to be.

✓ You're on the list. We'll be in touch.

How to compare LLM API prices honestly

Related: Self-host vs cloud calculator · Local LLM VRAM calculator · Token counter

Why the cheapest model changes with your usage

Every provider charges two prices: one for the tokens you send (input) and a higher one for the tokens the model writes back (output). A summarizer with long prompts and short answers is input-heavy; a chatbot that writes essays is output-heavy. Because the two rates differ by up to 6×, the cheapest model for a summarizer may not be the cheapest for a writer. This tool weights both by your actual mix and re-ranks the field live.

What is "effective cost per 1M tokens"?

It's the single number that matters: (input tokens × input price + output tokens × output price) ÷ total tokens. It collapses the two headline prices into what you'd truly pay per million tokens at your ratio — the honest apples-to-apples figure for comparing vendors.

When does owning hardware beat renting an API?

Local inference costs nothing per token — only electricity — against a fixed hardware cost up front. Past a break-even monthly volume, owning a GPU and running an open model wins, and the savings compound. The cost-smart pattern is to run everything you can locally and escalate to a paid API only when a local model misses your quality bar, which is exactly what ABUZ8 OS automates with cost-aware routing. See the break-even calculator for your numbers.

Is my data sent anywhere?

No. This calculator runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.