Where Sume documents plans, wallet and rates: which page answers what
Eight billing questions mapped to the docs page that answers each: the wallet, top-ups, list x 1.25, plan concurrency, 402 vs 429 and refunds, read 2026-10-09.

Sume documents billing across five docs pages, and each answers a different question: Billing and credits for the wallet and top-ups, Usage for the ledger, Generation admission for plan limits, Errors and rate limits for the status codes, and the model pages for the list x 1.25 rule. The docs give you the mechanics, not a plan price list.
That last point matters when you plan a budget. The docs say the API pricing page on sume.com is generated from the same pricing constants as the API bills, so use that page or the model catalog for rates, and use the docs for how money moves.
Question to page
Read 2026-10-09 from the repo docs on origin/main.
| Question | Docs page | What it says |
|---|---|---|
| What does Sume bill, and by how much? | Video generation; Video Router | Reserves provider list x 1.25 at submit; the poll response usage.cost is the Sume billable amount; pricing_skus per model in GET /v1/videos/models |
| Where do I add funds? | Billing and credits | Dashboard Billing & subscription; manual Stripe-backed top-up when billing is configured; no public API call creates a top-up |
| How do I read my balance? | Billing and credits; Usage | GET /v1/balance (USD; compatibility fields round to cents as credits) and GET /v1/usage |
| What did one job, run or thread cost? | Usage | GET /v1/usage with job_id, run_id or thread_id; quote summary.debited_usd |
| What are the plan limits? | Generation admission | Processing concurrency and queue capacity per plan; concurrency is plan-only |
| What does a 402 or 429 mean? | Generation admission; Errors and rate limits | 402 insufficient_credits is a balance problem; 429 queue_full and rate_limited are capacity problems |
| What is refunded? | Usage; Generation admission | Failed or canceled jobs release the reservation; cancel works only before generation starts |
| What do images and music cost? | Image API; Music Router | Per-model image lists and quality tiers are in the Image API page; music is selected through the Music Router catalog |
A five-line check of the whole flow
The fastest way to learn the model is to run one job. Read the balance, submit a 10-second Wan 3.0 720p clip (price $1.25), and read the usage rows for that job: you should see a reserved row, then a captured row of $1.25 after completion.
import os, requests
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
r = requests.get("https://api.sume.com/v1/balance", headers=H, timeout=30)
r.raise_for_status()
micros = r.json()["data"]["balance"]["available_amount_usd_micros"]
print(f"balance: ${micros / 1_000_000:.2f}")
# then submit one job and read its ledger rows:
# GET /v1/usage?job_id=<id> -> reserved, then capturedWhat the docs deliberately do not say
They do not publish how much a plan costs per month, and they do not list a per-model price table in prose. The usage page says the metered rates live on the API pricing page. They also warn against inventing prepaid top-ups through the API: top-ups are a dashboard operation.
Trust order for a number: the live catalog (pricing_skus), then the pricing page, then a blog table like the ones here. A blog table has a read date for a reason.
Gotchas
- Older docs and links call Billing & subscription 'Credits'; the product nav label is Billing & subscription.
- Concurrency comes from the plan; a top-up does not change it.
- The ledger shows reservations, captures, refunds, top-ups and grants, and the entries change with the workspace and environment.
Sources
Related posts
More in Pricing
- Why a 5-second Wan 3.0 480p job reserves 32 cents on Sume
Sume reserves the estimated price at submit and rounds up to a whole cent: 5 x $0.0625 = $0.3125 becomes 32 cents. Six worked examples from the catalog.
- Why Seedance 2.5 costs 4.6x Wan 3.0 per second at 720p
At 720p Seedance 2.5 is $0.578 per second on Sume and Wan 3.0 is $0.125. The gap comes from billing by video tokens (21,600 per second) versus a flat rate.
- Does audio off make AI video cheaper? Kling 3 yes, Wan 3.0 no
Turning generate_audio off cuts Kling 3 by a third on Sume but changes nothing on Wan 3.0, Seedance 2.5 or Omni. Which rows price sound, with a 5 s example.
- Haiku 5.5 with US-only inference: 1.1x on both price tiers
Anthropic's pricing page applies the 1.1x US-only multiplier to Haiku 5.5's over-100k prices too. A worked agent-loop table and where Sume's own cap sits.
Written by Sume