Sample and coding
The sample is every record labeled Data and AI infrastructure in the 201-company corpus. Researchers reviewed those records from 3 to 9 July 2026. We coded the full pricing-model field for self-serve entry, sales-assisted entry, numeric public prices, usage or consumption language, and private deployment terms. Source publication dates vary and are not replaced by the record review date.
Thirteen records contained both self-serve and sales-assisted signals, one contained self-serve without a matched sales signal, five contained sales language without a matched self-serve signal, and one contained neither. Negative statements such as “no public self-serve pricing found” were not counted as self-serve. These are text classifications, not measured revenue-channel shares.
What buyers can establish publicly
| Question | Records | What the signal establishes |
|---|---|---|
| Can a user start without a seller? | 14 / 20 | The record names a free, trial, open-source, or pay-as-you-go route |
| Does a larger configuration require sales? | 18 / 20 | The record names a quote, demo, custom, or sales-assisted route |
| Is a billing unit visible? | 12 / 20 | The record names usage, credits, tokens, compute time, data, or another consumption unit |
| Is any numeric product price visible? | 7 / 20 | At least one product currency amount or per-period figure appears; funding amounts are excluded |
| Is private deployment described? | 8 / 20 | The record names an evidenced self-hosted, on-premises, BYOC, VPC, hybrid, sovereign, or dedicated deployment |
Current pricing examples
Baseten's page checked on 8 September lists a $0-per-month pay-as-you-go Basic plan, per-token Model APIs, per-minute dedicated compute, quote-led Pro and Enterprise plans, and cloud, customer-infrastructure, and hybrid choices. A buyer can model a test workload publicly, but volume discounts and support levels still require a quote.
Qdrant's current page offers a free tier, usage-based Standard, a Premium tier with a minimum spend, and sales-assisted Hybrid and Private Cloud. Its tier table names SSO, private links, support, and SLA differences, while the actual Premium minimum is not shown on the public page.
Sources: Baseten cloud pricing · Qdrant cloud pricing
A practical comparison request
- Define the billable object: tokens, compute minutes, credits, storage, requests, data volume, or a combination.
- Model idle time, minimum spend, overages, data transfer, storage, and support separately from the headline unit rate.
- State where data and compute run, who operates the deployment, and which commitments or infrastructure costs remain with the customer.
- Request the effective unit price at low, expected, and peak workloads. Public starting prices do not establish the contracted rate.
Limits
The cohort is a descriptive collection, not a market sample. Matching terms can miss unusual language. Numeric-price presence does not mean a full production price is public, and a sales signal does not establish that every purchase requires a seller.
Sources and dates
- Baseten cloud pricing ↗
Current check for plan, unit, support, and deployment examples.
Reviewed 2026-09-08 - Qdrant cloud pricing ↗
Current check for tier, minimum-spend, support, SLA, and deployment examples.
Reviewed 2026-09-08 - Contextual AI pricing ↗
First-party pricing source cited in the cohort record.
Reviewed 2026-07-03 - Firebolt pricing ↗
First-party pricing source cited in the cohort record.
Reviewed 2026-07-09
Found an error or a changed source? Send a correction.
Apply this to your company
Prepare materials for your next buyer conversation.
Accountmade helps technical B2B teams prepare demo decks, technical blueprints, business cases and security materials.
Explore Accountmade →