(Free tool)AI cost checkFor founders and product leads
What does your AI feature cost per user?
What one active user costs you in model fees each month, against their seat price. From list prices read on 19 October 2026.
Runs in your browser. Nothing you type leaves the page.
Open source, MIT: the code on GitHub.
A user, a month · Claude Sonnet 5.5
US$0.44
US$0.011 an action × 40 actions
2.2% of a US$20 seat
Past the seat: tangled again.
- Fills the seat at
- 1,818 actions
- The frame
- 1.1 of 48 lines
- Left after model fees
- 97.8%
US$19.56 of a US$20 seat left after the model
At these figures an action costs nothing, so nothing pours into the seat.
- A user, a month
- US$0.44 · 2.2% of a US$20 seat
- AI cost check
- Cost against the seat
- Prices read
- 19.10.2026
- The seat
- US$20
Every model, at your numbers.
The same month on each model on the list, small to large. Your choice is lit; the bars share one scale.
| Model | An action | A user, a month | Share of the seat | Break-even, actions |
|---|---|---|---|---|
| Anthropic read 19.10.2026 | ||||
| Claude Haiku 5.5 small | US$0.00055 | US$0.022 | 0.1% | 36,363 |
| Claude Sonnet 5.5 mid your choice | US$0.011 | US$0.44 | 2.2% | 1,818 |
| Claude Opus 5.5 large | US$0.022 | US$0.88 | 4.4% | 909 |
| Claude Fable 5.1 large | US$0.055 | US$2.20 | 11.0% | 363 |
| Google read 19.10.2026 | ||||
| Gemini 3.5 Flash-Lite small | US$0.00215 | US$0.086 | 0.4% | 9,302 |
| Gemini 3.8 Flash mid new price from 1 January 2027 | US$0.00413 | US$0.165 | 0.8% | 4,848 |
| Gemini 4 Argon large new price from 1 December 2026 | US$0.011 | US$0.44 | 2.2% | 1,818 |
| OpenAI read 19.10.2026 | ||||
| GPT-6 Luna small | US$0.00055 | US$0.022 | 0.1% | 36,363 |
| GPT-6.1 Sol mid | US$0.011 | US$0.44 | 2.2% | 1,818 |
| GPT-6 Astra large | US$0.055 | US$2.20 | 11.0% | 363 |
Three parts, one month.
Each part is tokens times its rate. The cached share is billed at the cache-read price, the rest at the input price.
As use grows, at three tiers
The chart as a table
| Actions a month | Claude Haiku 5.5 | Claude Sonnet 5.5 | Claude Opus 5.5 |
|---|---|---|---|
| 0 | US$0.00 | US$0.00 | US$0.00 |
| 40 (yours) | US$0.022 | US$0.44 | US$0.88 |
| 500 | US$0.275 | US$5.50 | US$11.00 |
| 1,000 | US$0.55 | US$11.00 | US$22.00 |
| 1,500 | US$0.825 | US$16.50 | US$33.00 |
| 2,000 | US$1.10 | US$22.00 | US$44.00 |
| Part | Tokens a month | Rate, per million | Cost |
|---|---|---|---|
| Input 54.5% | 120,000 | US$2.00 | US$0.24 |
| Cached 0.0% | 0 | US$0.10 | US$0.00 |
| Output 45.5% | 20,000 | US$10.00 | US$0.20 |
| A user, a month | US$0.44 | ||
The opening example, on Claude Sonnet 5.5: 40 actions × 3,000 tokens is 120,000 input tokens, US$0.24; 40 × 500 is 20,000 output tokens, US$0.20. Together, US$0.44 a user a month. Break-even is the US$20 seat divided by US$0.011 an action: 1,818 actions.
What the sum leaves out
- Cache writes. Anthropic charges 1.25× the input price for a five-minute cache write, 2× for an hour (1). OpenAI lists a write price (5); Google bills storage hourly (4).
- Discounts and uplifts. Anthropic’s half-price batch rate, a negotiated discount, or the 10% uplift Anthropic and OpenAI add for one-region processing (1, 5). For these, choose “Your own”.
- Tools and retries. Web search and other tools are priced on their own. A retried call is billed again: count retries as actions.
- Everything else. Hosting, storage, payment fees, support and tax. The margin here is after AI cost only, not your gross margin.
- Tokens per model. The same words can be more tokens on a newer model: Anthropic’s newer tokenizer “produces approximately 30% more tokens for the same text” (1, 2). Thinking counts as output (3, 4).
The prices, dated and sourced.
The list on one scale
Standard rates from each provider’s own page on 19 October 2026. Small, mid and large are this page’s words, not theirs.
The price list, read 19 Oct 2026
| Model | Tier | Input | Cache read | Output | Notes |
|---|---|---|---|---|---|
| Anthropic Pricing read 19.10.2026 | |||||
Claude Haiku 5.5claude-haiku-5-5 | small | US$0.10 | US$0.01 | US$0.50 | Over 100,000 input tokens a request: US$0.50 in, US$0.05 cached and US$2.50 out. |
Claude Sonnet 5.5claude-sonnet-5-5 | mid | US$2.00 | US$0.10 | US$10.00 | A cache read is 0.05 times the input price on this model. |
Claude Opus 5.5claude-opus-5-5 | large | US$4.00 | US$0.20 | US$20.00 | A cache read is 0.05 times the input price on this model. |
Claude Fable 5.1claude-fable-5-1 | large | US$10.00 | US$0.25 | US$50.00 | A cache read is 0.025 times the input price on this model. |
| Google Gemini Developer API pricing read 19.10.2026 · page updated 16.10.2026 | |||||
Gemini 3.5 Flash-Litegemini-3.5-flash-lite | small | US$0.30 | US$0.03 | US$2.50 | An explicit cache also pays storage: US$1.00 per million tokens an hour. |
Gemini 3.8 Flashgemini-3.8-flash | mid | US$0.75 | US$0.075 | US$3.75 | Until 31 December 2026. From 1 January 2027: US$1.50 in, US$0.15 cached and US$7.50 out. An explicit cache also pays storage: US$0.50 per million tokens an hour, US$1.00 from 1 January 2027. |
Gemini 4 Argongemini-4-argon | large | US$2.00 | US$0.10 | US$10.00 | Until 30 November 2026. From 1 December 2026: US$4.00 in, US$0.20 cached and US$20.00 out. Implicit caching only. |
| OpenAI API pricing read 19.10.2026 | |||||
GPT-6 Lunagpt-6-luna | small | US$0.10 | US$0.01 | US$0.50 | Over 272,000 input tokens a request: US$0.20 in, US$0.02 cached and US$0.75 out. |
GPT-6.1 Solgpt-6.1-sol | mid | US$2.00 | US$0.10 | US$10.00 | Over 272,000 input tokens a request: US$4.00 in, US$0.20 cached and US$15.00 out. |
GPT-6 Astragpt-6-astra | large | US$10.00 | US$1.00 | US$50.00 | Over 272,000 input tokens a request: US$20.00 in, US$2.00 cached and US$75.00 out. |
Sources
- Anthropic, “Pricing”, platform.claude.com/docs/en/about-claude/pricing, read on 19 October 2026: model prices, cache writes and reads, Claude Haiku 5.5’s prices over 100,000 input tokens, the batch discount, the US-only processing multiplier and the tokenizer note.
- Anthropic, “Models overview”, platform.claude.com/docs/en/about-claude/models/overview, read on 19 October 2026: the current models, and the tokenizer introduced with Claude Opus 4.7.
- Anthropic, “Extended thinking”, platform.claude.com/docs/en/build-with-claude/extended-thinking, read on 19 October 2026: thinking tokens are reported as part of the billed output tokens.
- Google, “Gemini Developer API pricing”, ai.google.dev/gemini-api/docs/pricing, last updated 16 October 2026, read on 19 October 2026: paid-tier standard prices, the dated changes, context caching and its storage, and output prices that include thinking tokens.
- OpenAI, “API pricing”, developers.openai.com/api/docs/pricing, read on 19 October 2026: standard-tier text-token prices, for requests of up to 272,000 input tokens and for longer ones, cache writes, and the 10% uplift for regional processing.
Not affiliated with any provider named here. The check gives an estimate from your own figures and list prices; it is not a quote, and not financial advice. The code, this price list and its tests are open source, MIT: github.com/growmint-media/ai-cost-check.
When the number has to hold up.
This check runs on list prices and your estimates. The audit works from the feature itself, live or planned: its jobs, failure modes, test cases and cost against your price.
Start to
Have the whole feature auditedUS$4,500
Most clients start with a teardown. The Teardown, US$950
- AI cost check
- Result
- Date
- Today
- Prices read
- 19.10.2026
A user, a month · Claude Sonnet 5.5
US$0.44
2.2% of the seat · fills the seat at 1,818 actions · left after model fees 97.8%
- Model
- Claude Sonnet 5.5
- Actions a user, a month
- 40
- Input tokens an action
- 3,000
- Output tokens an action
- 500
- Cached share of the input
- 0%
- Seat price a user, a month
- US$20
- Rates, per million tokens
- US$2.00 in · US$0.10 cached · US$10.00 out
- Prices
- Anthropic’s list, read 19 October 2026
An estimate from your figures and list prices, not a quote. These numbers: quietlybuild.in/tools/ai-cost-check
Have the whole feature audited · quietlybuild.in/studio/ai-audit · US$4,500