Research · Synthesis
AI Pricing: Tokens, Credits or Outcomes?
Tokens tie an application’s price to model consumption; credits can package different units, so what triggers a charge matters. Tugce Erten and Sarah Wang recommend pricing the highest value layer that can be measured, attributed and defended. Mintlify kept credits but moved from token-variable charges to fixed prices for answers and document updates, with specified no-result cases free. These sources do not establish one model for every AI product.
Saves the question and link in this browser. Reopening reads the current reviewed version.
Lines show the reading structure. Select an idea to read its explanation and evidence.
Tokens: cost is not the whole value
Tugce Erten and Sarah Wang argue that per-token application pricing imports the model provider’s cost structure and anchors value to a unit whose cost keeps falling. They describe carrying model-layer pricing into the application layer as often a mistake; this is their pricing judgment, rather than a universal prohibition.
Supporting evidence
Original excerpt
When an application prices that work in tokens, it imports the model provider’s cost structure into the relationship with the customer and anchors the product’s value to a unit whose cost keeps falling. Based on our work, it’s often a mistake to carry the model layer’s pricing logic into the application layer.
Credits: inspect the charge trigger
Han Wang says Mintlify’s credits previously varied with the tokens used in each run. The announcement changes this to fixed charges: 25 credits for an assistant answer and 250 for a document update from an automation run. Credits remain the unit, while the billable event changes.
Supporting evidence
Original excerpt
Now pricing reflects the value we visibly provide. An assistant answer is 25 credits. A docs update from an automation run costs 250. If the Assistant can't answer, or an automation determines nothing needs updating, you pay nothing.
Original excerpt
Credits previously varied with the tokens used by each run. Now every action has a fixed price, and actions that produce nothing cost nothing:
Failures: define the no-charge cases
Han Wang specifies that an assistant unable to answer, or an automation determining that nothing needs updating, costs nothing. These are the stated no-result cases. The cited passage does not specify settlement for every technical failure, a user-requested retry or a disputed answer.
Supporting evidence
Original excerpt
Now pricing reflects the value we visibly provide. An assistant answer is 25 credits. A docs update from an automation run costs 250. If the Assistant can't answer, or an automation determines nothing needs updating, you pay nothing.
Predictability: separate the event from the plan
Han Wang gives fixed prices for answers and automation updates, while saying plans, included credits and the price per credit remain unchanged. Editor and Slack agent usage becomes unlimited on Pro and Enterprise plans. This lets readers distinguish the stated price of an event from what a plan includes; the excerpts do not establish a customer’s total bill or complete dispute records.
Supporting evidence
Original excerpt
Now pricing reflects the value we visibly provide. An assistant answer is 25 credits. A docs update from an automation run costs 250. If the Assistant can't answer, or an automation determines nothing needs updating, you pay nothing.
Original excerpt
We are also providing unlimited editor agent and Slack agent usage with Pro and Enterprise plans. Your plan, included credits, and the price per credit remain unchanged.
Outcomes: require defensible value
Tugce Erten and Sarah Wang recommend the highest value layer that can reliably be measured, attributed and defended; they describe credits as packaging understandable units and moving toward outcomes when customers can recognize and trust them. Mintlify’s named results are answers and document updates. The cited passages do not establish downstream business success, delivery costs or margins.
Supporting evidence
Original excerpt
Now pricing reflects the value we visibly provide. An assistant answer is 25 credits. A docs update from an automation run costs 250. If the Assistant can't answer, or an automation determines nothing needs updating, you pay nothing.
Original excerpt
Instead, we believe companies should price at the highest layer of value that they can reliably measure, attribute, and defend.
Original excerpt
The better path is to price at the highest layer of value you can reliably measure, attribute, and defend. Translate variable work into understandable units. Use credits to package those units when flexibility matters. Move toward outcomes as soon as customers can recognize and trust them.
Scope and limitations
- This synthesis compares a16z’s August 27, 2026 pricing framework with Mintlify’s September 8, 2026 announcement. They address different tasks and are not presented as a direct debate. Prices describe that announcement, not a continuously verified live tariff.
- Mintlify’s two explicit no-charge conditions do not settle every error, retry, cancellation or disputed result. Its separate billing documentation is not part of the admitted evidence used here.
- For a pricing decision, the open questions are the billable event, failure and retry settlement, advance predictability, dispute evidence and who bears delivery costs. The excerpts answer some of these and leave others unresolved; no margin or universal contract is inferred.
- The a16z statements retain the joint attribution to Tugce Erten and Sarah Wang. The Mintlify statements are attributed to Han Wang’s announcement. This is a reviewed editorial synthesis, not a product test or an industry-wide statistic.
These statements reflect the dates and contexts of the cited sources. Differences in scope do not establish disagreement or a change of position.
Bring your own question
Use this question as a starting point. Edit it before searching across the available collection.
Prepare your question →Sources and review
Evidence at a glance
Evidence at a glance
Prepared by nafyi with AI assistance and a separate semantic verification pass against approved source evidence. Check important conclusions in the original sources.
Version 1 · Updated · Semantic review