--- title: AI Pricing Finder, how 96 products price AI description: Component-level payment timing, billing terms, overage, rollover and top-ups for 96 live products, with the source text behind every established answer. date: 2026-09-03 --- > Full site context for agents: https://tansohq.com/llms.txt # AI Pricing Finder How 96 products price AI. The comparison shows one row per current product; the downloadable data preserves all 410 pricing components. Last updated September 4, 2026. ## The six fields - **Pricing metric.** What units, balances, quantities or fees affect the price? The `main_metric` flag identifies the component expected to drive most customer spending. The separate `answers_row` flag identifies the component highlighted for the named AI feature: a standard included allowance first, then a customer-funded balance, a contracted amount, or the main post-use meter. A product-specific add-on is used only when it is the published measurable mechanism governing additional AI use. - **Commitment timing.** Does the customer buy or commit to that component before using it, pay after using it, or do both? The values are `Before use`, `After use` and `Both`. This describes the order of commitment and use, not the calendar date on which an invoice is paid: a committed quantity can still be invoiced in arrears. A seat can use any of these paths; its name does not decide. - **Billing term.** What period applies to that component: Daily, Weekly, Monthly, Quarterly, Semiannual, Annual, Multi-year or None? - **Overage.** Can prepaid usage be exceeded and billed? - **Rollover.** Do unused credits carry at the end of the term? Yes if any amount is kept for any time after the term. With no recurring term, Yes or No requires a documented link to another plan or cycle; otherwise Rollover is N/A because no boundary exists. - **Top up.** Can the customer buy more of the priced component during the current billing cycle without changing plans? This includes credits, usage packs, capacity, and seats. Like the other four fields it is answered for the highlighted component, but the money may arrive by any route: the answer is Yes whether it adds to the same balance, buys a separate prepaid balance, or funds a documented add-on or companion product, provided the purchase buys usage for the component being scored. Where the mechanism is a separate balance the comparison marks it Yes°, and the product's detail panel names it. - **Commitment term.** What is the shortest period a self-serve customer must commit to for the product's main paid offering: none, monthly, quarterly, annual or multi-year? A metered API that can be abandoned at any time is `none`. Where a vendor sells both monthly and annual, the field records the shorter. - **Commitment term.** What is the shortest period a self-serve customer must commit to for the product's main paid offering: none, monthly, quarterly, annual or multi-year? A metered API that can be abandoned at any time is `none`. Where a vendor sells both monthly and annual, the field records the shorter. Values are Yes, No, N/A or a dash. N/A applies when a component is metered and billed entirely after use, and to Rollover when no applicable term boundary exists. A dash means the evidence does not establish an answer; absence is never treated as No. Every Yes or No rests on quoted source text with its URL and read date, shown on each product's page (/ai-pricing-finder/{id}). ## Rows are products OpenAI's API and ChatGPT are separate rows, as are Anthropic's API and Claude. 5 products are retired, not deleted, so published citations still resolve: they stay in the export with the reason and keep their pages. They are not listed in the table. ## Data - [Current export](https://tansohq.com/data/pricing-finder-current.json): all 101 products, every component and every quote - [Frozen September 1 snapshot](https://tansohq.com/data/pricing-finder-snapshot-2026-09-01.json): the dataset before the September rebuild - [Frozen August 30 snapshot](https://tansohq.com/data/pricing-finder-snapshot-2026-08-30.json): superseded - [Original June 17 snapshot](https://tansohq.com/data/pricing-finder-snapshot-2026-06-17.json): the 73-company set analysed in "Pricing for the Agent Economy" - [Revised June 17 snapshot](https://tansohq.com/data/pricing-finder-snapshot-2026-06-17-v2.json) The page also offers a CSV download: one row per pricing component, with separate `main_metric` and `answers_row` flags and every answer, quote and source URL. ## Provenance Each evidence field carries three columns, not one, so a reader can judge a value without opening the URL: - `_confidence` — `high` for first-party billing or contractual documentation, `medium` for first-party pricing or marketing language or a documented inference, `low` for secondary evidence or a source of uncertain date. Blank for `?` values. - `_source` — `vendor` where the quote is the vendor's own words, `third party` otherwise. The label follows the company rather than the hostname, because a vendor often documents elsewhere: Segment's docs are on twilio.com, Loom's on atlassian.com, Slack's Agentforce terms on salesforce.com. - `_quote` and `_url` — the sentence and the page it was read on. Where no source states an answer but the vendor's own text constrains it, the value is a documented inference and its note begins "INFERRED, not stated by the vendor", followed by what the inference rests on. Competitor comparison pages are not treated as evidence at any confidence: where an oddly specific figure appears across several such pages with no origin, it is one unsourced claim rather than several sources, and it is left out with the reason recorded. ## Links - [HTML version](https://tansohq.com/ai-pricing-finder) - [Methodology](https://tansohq.com/methodology.md) More for agents: https://tansohq.com/llms.txt