# Baseten

> baseten.co | Compute, Storage; 3 records: free tier, limits, identity cost and evidence.

HTML https://fshy.ajam.dev/p/baseten/ | JSON https://fshy.ajam.dev/api/v1/entity/baseten.json | how to read a record: https://fshy.ajam.dev/llms-full.txt

## Baseten (published tiers, no free plan)

- `compute/baseten` | paid-only | evidence documented | binding none | category GPU cloud | tier published tiers, no free plan
- needs card yes, phone no, ID no, institution no | verified 2026-10-24T01:52:00.000Z (live) | stale on 2026-11-24
- url https://www.baseten.co/pricing/
- free tier: no free tier on the pricing page: 'Only pay for the time your model is actively deploying, scaling up or down, or making predictions.' | unit: none | metered on: none | read on: 2026-10-24 | source: [1] | note: The word 'free' does not appear on the page.
- per-object cap: not applicable: the unit is a session, not an object
- retention: not applicable: no free allowance | source: [1]
- egress: not stated | metered: no | source: [1]
- identity: card | detail: a paid account: the pricing page describes only pay-for-use billing and states no free plan
- interface: unknown
- limits:
  - volume-quota: 0 hours: no free tier published [1]
- limits note: not stated
- binding note: there is no free allowance to exhaust
- tooling: the same as limits note
- composes with: the same as limits note
- reliability: no reliability event recorded
- notes: Baseten is a serving platform, not a notebook host, so it never had hours to give. It is recorded because provider list names it.
- evidence note: the absence of a free tier is read from the full pricing page.
- extras:
  - keep_rate: 1.0
  - billing_model: usage
  - session_cap.restart_required: no
  - session_cap.note: not applicable
  - free_tier_kind: none
  - granularity.min_billed_s: 60
  - term: monthly
  - price_arithmetic: not published in a comparable hourly unit
  - units: RAM and disk are quoted in the vendor's own decimal GB, not GiB. The ram_gib and disk_gib keys carry the vendor's number unconverted, so where a vendor writes 30 GB the key holds 30 and the true value is about 27.9 GiB. No conversion was performed.
- evidence:
  - [2] 2026-10-02: 
  - [2] 2026-10-22: the resources documentation was read on 2026-10-22 and describes instance selection, vCPU/RAM/GPU sizing and multi-node inference; it publishes no free tier and no rate
  - [1] 2026-10-24: quoted: 'Only pay for the time your model is actively deploying, scaling up or down, or making predictions.'
  - [1] 2026-10-24: the page contains no free-tier row and no occurrence of 'free'
- sources:
  - [1] https://www.baseten.co/pricing/
  - [2] https://docs.baseten.co/deployment/resources

## baseten (Storage)

- `storage/baseten` | paid-only | evidence reported | binding none | category rejected-provider
- identity a placeholder: card and phone unknown | verified 2026-10-21 (inherited) | stale on 2026-11-05
- url https://www.baseten.co/pricing/
- free tier: Inference-only PaaS. The pricing page is a model-API price list ($/token) plus per-second deployment billing; there is no storage product, no artifact store, and no free tier bytes to claim. | unit: none | metered on: none | read on: 2026-10-21 | source: [1]
- per-object cap: not recorded | source: [1] | note: not recorded
- retention: not recorded | source: [1]
- egress: not recorded | metered: no
- identity: email
- interface: unknown
- tooling: not recorded
- composes with: the same as tooling
- reliability: the same as the free tier
- notes: the same as the free tier
- extras:
  - dead_end.failure_class: paid
  - dead_end.why: Inference-only PaaS. The pricing page is a model-API price list ($/token) plus per-second deployment billing; there is no storage product, no artifact store, and no free tier bytes to claim.
  - dead_end.first_recorded: 2026-10-21
- evidence:
  - [1] 2026-10-21: Inference-only PaaS. The pricing page is a model-API price list ($/token) plus per-second deployment billing; there is no storage product, no artifact store, and no free tier bytes to claim.
- sources:
  - [1] https://www.baseten.co/pricing/

## Baseten Delivery Network (BDN) model weight mirrors

- `storage/baseten-delivery-network-model-weight-mirrors` | fails-egress | evidence reported | binding none | category paas
- identity a placeholder: card and phone unknown | verified 2026-10-21 (inherited) | stale on 2026-11-05
- url https://baseten.co
- free tier: No published per-account storage number. BDN 'mirrors model weights to Baseten and caches them near your replicas'; there is no documented quota and no free tier stated. Sized for 'a few GB or less' (weights that can be bundled directly with a Truss) up to multi-GB weight sources. | unit: GB | metered on: stored-bytes | read on: 2026-10-21 | source: [1]
- per-object cap: not stated | source: [1] | note: not stated
- retention: not stated | source: [1] | note: not stated
- egress: Not applicable — BDN is a read-only cache mounted into your own replicas ('mounted read-only'), not a public download surface. No anonymous URL. | metered: no
- identity: anonymous
- interface: rest
- limits note: per-object-cap: not published; the system expects weights in the GB range, not the 100 GB+ range. Retention: cached by config hash, evicted by the platform ('Caching and deduplication happen behind the scenes'). No free allowance stated.
- interface detail: rest
- tooling: `baseten model push` (Truss), management API POST /v1/prepare_model_upload ('Validates a model push payload and issues upload credentials'), config.yaml `weights:` / `bdn:` keys.
- composes with: Not a storage product for a user. It is a private artifact cache inside a paid inference platform.
- reliability: Requires an organisation account and a paid inference deployment to exist before weights are cached; withdrawing the deployment removes the mount.
- notes: the same as the free tier
- extras:
  - dead_end.failure_class: quota
  - dead_end.why: No published per-account storage number. BDN 'mirrors model weights to Baseten and caches them near your replicas'; there is no documented quota and no free tier stated. Sized for 'a few GB or less' (weights that can be bundled directly with a Truss) up to multi-GB weight sources.
  - dead_end.first_recorded: 2026-10-21
- evidence:
  - [1] 2026-10-21: 'Mirror and cache model weights ... mounted read-only'; 'If your model is small (a few GB or less) you can also bundle weights directly with your Truss'
  - [2] 2026-10-21: POST /v1/prepare_model_upload 'Validates a model push payload and issues upload credentials'
- sources:
  - [1] https://docs.baseten.co/development/model/bdn
  - [2] https://docs.baseten.co/reference/management-api/models/prepare-model-upload