GPU Boosts

GPU Boosts give you a reserved slice of a shared GPU pool — guaranteed tokens/sec and RPM at a fixed monthly price. Each pool runs a single model; you pick which time blocks you want and how many seats.

All endpoints require authentication. Use your management key (mk_):

Authorization: Bearer mk_your_key

---

List available pools

GET /api/pools

Returns all pools that are not in draft status.

curl https://api.cheapestinference.com/api/pools \
  -H "Authorization: Bearer mk_your_key"

{
  "success": true,
  "data": [
    {
      "id": "pool_uuid",
      "slug": "qwen3-397b-a17b",
      "modelId": "Qwen/Qwen3.5-397B-A17B",
      "modelName": "Qwen3.5-397B-A17B",
      "description": "Dedicated Qwen3.5 inference at guaranteed throughput.",
      "infraSpec": "8× H100 SXM5 · 50 tok/s guaranteed",
      "status": "funding",
      "totalSlots": 10,
      "pledgedSlots": 3,
      "minPricePerDay": "27.00"
    }
  ]
}

Field	Description
status	funding (accepting pledges), activating, active, paused
totalSlots	Maximum seats available across all blocks
pledgedSlots	Number of confirmed reservations
minPricePerDay	Lowest monthly price across all blocks (USD)

---

Pool detail

GET /api/pools/:id

Returns full pool info including all hour slots. :id can be the UUID or slug.

curl https://api.cheapestinference.com/api/pools/pool_uuid \
  -H "Authorization: Bearer mk_your_key"

{
  "success": true,
  "data": {
    "id": "pool_uuid",
    "slug": "qwen3-397b-a17b",
    "modelId": "Qwen/Qwen3.5-397B-A17B",
    "status": "funding",
    "hourSlots": [
      { "id": "slot_uuid", "hour": 0, "slotIndex": 0, "pricePerDay": "3.38", "status": "available" },
      { "id": "slot_uuid", "hour": 1, "slotIndex": 0, "pricePerDay": "3.37", "status": "available" }
    ]
  }
}

Each hourSlot represents one hour of one seat. Prices are monthly amounts distributed across the 8 hours in a block.

---

Subscribe (recommended)

POST /api/pools/:id/subscribe

Reserve seats using block names instead of raw slot IDs. This is the preferred API — it handles slot selection automatically and enforces block-level granularity.

Time blocks

Each pool day is divided into three fixed UTC blocks:

Block	Hours (UTC)	Typical coverage
night	00:00–07:59	Asia-Pacific
europe	08:00–15:59	Europe / Middle East
americas	16:00–23:59	Americas

Request body

Parameter	Type	Required	Description
blocks	string[]	Yes	One or more block names: "night", "europe", "americas"
quantity	integer	No	Number of seats per block (1–20, default 1)

Example — single seat, one block

curl -X POST https://api.cheapestinference.com/api/pools/pool_uuid/subscribe \
  -H "Authorization: Bearer mk_your_key" \
  -H "Content-Type: application/json" \
  -d '{"blocks": ["americas"]}'

Example — 2 seats across all blocks

curl -X POST https://api.cheapestinference.com/api/pools/pool_uuid/subscribe \
  -H "Authorization: Bearer mk_your_key" \
  -H "Content-Type: application/json" \
  -d '{"blocks": ["night", "europe", "americas"], "quantity": 2}'

Response

{
  "success": true,
  "data": {
    "id": "pledge_uuid",
    "poolId": "pool_uuid",
    "status": "pledged",
    "monthlyPrice": "27.00",
    "slotCount": 8
  }
}

If the pool is already active, the subscription activates immediately and status will be active with a key field included.

Pricing

monthlyPrice = sum of prices for all reserved slots. Each block has a fixed monthly price set by the pool operator. With quantity: 2 and two blocks selected, you pay 2 × (block_a_price + block_b_price).

What happens after reserving

• A $1 non-refundable reservation fee is charged to your card on file to confirm your payment method.
• The subscription activates when the pool reaches its minimum seat threshold (funding → active).
• Once active, your monthly charge is billed automatically and you receive a dedicated API key.

---

View my reservation

GET /api/pools/:id/my-pledge

curl https://api.cheapestinference.com/api/pools/pool_uuid/my-pledge \
  -H "Authorization: Bearer mk_your_key"

{
  "success": true,
  "data": {
    "id": "pledge_uuid",
    "status": "active",
    "monthlyPrice": "27.00",
    "currentPeriodEnd": "2026-05-04T00:00:00.000Z",
    "hours": [
      { "hour": 16, "slotIndex": 0, "pricePerDay": "3.38" }
    ],
    "key": {
      "id": "key_uuid",
      "apiKey": "sk_pool_abc123...",
      "rpmLimit": 60,
      "tpmLimit": 100000,
      "isActive": true
    }
  }
}

status	Meaning
pledged	Reserved, waiting for pool to activate
active	Subscription running, key is live
past_due	Payment failed, key suspended
canceled	Subscription ended

---

Get my pool key

GET /api/pools/:id/my-key

Returns only the key object. Useful if you just need the API key without the full pledge details.

curl https://api.cheapestinference.com/api/pools/pool_uuid/my-key \
  -H "Authorization: Bearer mk_your_key"

{
  "success": true,
  "data": {
    "id": "key_uuid",
    "apiKey": "sk_pool_abc123...",
    "rpmLimit": 60,
    "tpmLimit": 100000,
    "isActive": true
  }
}

Use apiKey as the Authorization: Bearer token on any /v1/* or /anthropic/* inference endpoint.

---

Cancel reservation

DELETE /api/pools/:id/pledge

Cancels a pledged (not yet active) reservation and frees your slots. Once a subscription is active, cancellation goes through the billing portal instead.

curl -X DELETE https://api.cheapestinference.com/api/pools/pool_uuid/pledge \
  -H "Authorization: Bearer mk_your_key"

{
  "success": true,
  "data": { "canceled": true }
}

---

How GPU Boosts work

1. Reserve — pick a pool, select time blocks, pay the $1 reservation fee. You’re committed to the first billing cycle.
2. Fund — the pool collects pledges until it hits its minimum seat count.
3. Activate — once fully funded, the pool goes live. All pledges are charged their monthly fee and receive a dedicated key.
4. Use — your key gets guaranteed RPM and tok/s during your reserved blocks. Use it exactly like any other API key.
5. Renew — subscriptions recur monthly automatically. Cancel anytime — access continues to the end of the paid period.