AI Token Cost Calculator
| Model | $/M in | $/M out | Estimated cost |
|---|---|---|---|
| Amazon Nova MicroAmazon | $0.035 | $0.14 | — |
| Claude Haiku 4.5Anthropic | $1 | $5 | — |
| GPT-5.2OpenAI | $1.75 | $14 | — |
| Claude Sonnet 5Anthropic | $2 | $10 | — |
| Gemini 3.1 ProGoogle | $2 | $12 | — |
| Claude Opus 5Anthropic | $5 | $25 | — |
| GPT-5.6 SolOpenAI | $5 | $30 | — |
| Claude Fable 5.1Anthropic | $10 | $50 | — |
Pricing snapshot from 2026-09-03, sourced from provider pricing pages — not fetched live, so double-check the provider's own pricing page before committing real spend, especially for high-volume use. Also note: the cheapest price-per-token isn't always the cheapest price-per-task — a model that needs more retries, more reasoning tokens, or longer outputs to get a usable answer can cost more overall than a nominally pricier one. Token counts from pasted text use a rough ≈4-characters-per-token estimate, not an exact tokenizer.
Estimate and compare API costs across 8 current AI models — paste a prompt or enter token counts directly, no sign-up.
How it works
- Paste your prompt (optional) to estimate input tokens, or enter token counts directly
- Enter your expected output token count
- See estimated cost across 8 models from Anthropic, OpenAI, Google, and Amazon, side by side
Frequently asked questions
How current is this pricing?
This is a point-in-time snapshot sourced from provider pricing pages, not a live feed — this site's tools run entirely in your browser and don't call third-party APIs at runtime. Provider prices can change without notice, so check the provider's own pricing page before committing real spend, especially for high-volume use.
Is the cheapest price-per-token always the cheapest option?
Not necessarily. A model with a lower listed price can still cost more per task if it needs more retries, more reasoning tokens, or longer outputs to produce a usable answer. Price-per-token is a starting point, not the full picture.
How accurate is the token estimate from pasted text?
It uses a rough ≈4-characters-per-token approximation, a commonly cited rule of thumb — not an exact tokenizer count, which varies slightly by model and provider.
Related tools
From the blog
Update history
- 2026-09-03 Launched: AI token cost calculator covering 8 current models across Anthropic, OpenAI, Google, and Amazon, with an optional paste-to-estimate input. Entirely client-side; pricing snapshot dated in the tool itself.