Seedream on BytePlus ModelArk

# Fix list: Seedream on BytePlus ModelArk

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/seedream-on-byteplus-modelark, the October 2026 research run, assessed 9 October 2026. Grade D, 53.8 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Seedream on BytePlus ModelArk: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 43 out of 100, up to 11.4 more on the total

Why it scored 43: Graded as a hosted service. No status page or incident history was found in the docs or the site pages read (0), so the 90-day record is scored as unreadable (5). The model list gives 500 images a minute for each Seedream model, with a note that limits are theoretical maximums and not guaranteed (15). The error reference lists 15 distinct 429 codes including `ModelAccountIpmRateLimitExceeded`, and a burst-traffic guide gives exponential backoff with jitter. No idempotency key was found (13). The legal index lists SLAs for 16 other BytePlus products and none named for ModelArk. We did not open those documents (0). The image API is not labelled beta or preview (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Payments & pricing, 20 out of 100, up to 10 more on the total

Why it scored 20: No x402, MPP or L402 (0). Per-image prices for every Seedream model are published without a login (20). No free image quota is listed for Seedream. The free trial page covers models billed by token, and the model list shows no free quota in the image table (0). Access needs a browser sign-up, model activation in the console and a console-made key (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 3. Schema & documentation, 62 out of 100, up to 6.2 more on the total

Why it scored 62: No OpenAPI file for the data plane was found. The Go and Python SDKs publish typed request and response models, and an API Explorer page exists that we did not read (8). No llms.txt or Markdown docs for agents were found, and the docs site is drawn by script (0). Each parameter states its purpose, the models that accept it and its restrictions (15). Inputs carry types, defaults, enums and pixel and aspect-ratio ranges, though `size` is a string with two exclusive formats (13). The tutorial has curl, Python, Java, Go and OpenAI SDK examples, and the error reference maps HTTP status to code, message and meaning (14). The path is versioned at `/api/v3`, model IDs carry a date, and release pages are dated by month (12).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 4. Security & auth, 70 out of 100, up to 5.3 more on the total

Why it scored 70: API keys are scoped to a resource project and region, can be limited to named model IDs or endpoints and to an IP allowlist, and can be disabled or deleted. No expiry or rotation feature was found. The key travels in a header (26). Model-level key permissions, per-endpoint rate limits and account inference limits narrow what a key can spend, with no read-only notion for a generation API (12). The API returns generated images only (10). The console shows each key's last use and creator, and a usage query API exists. The inference security audit covers only Skylark models (8). No security.txt. The SDK repositories send reports to ByteDance's security centre, which runs a bounty, and the trust centre lists ISO 27001, SOC 2 and PCI DSS for BytePlus as a whole (14).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 5. Agent ergonomics, 72 out of 100, up to 4.6 more on the total

Why it scored 72: Graded on the model lines. Responses can be a URL or `b64_json`, `size` takes resolution levels, and `max_images` caps a batch (20). There is nothing to paginate. Sequential generation, streaming and per-request output controls cover size (14). Errors carry a code, a message and a request ID, and a failed image in a batch reports its own error without failing the rest (17). No idempotency key was found. The SDK retries twice by default, and images that fail are not billed (6). Only `model` and a prompt or image are needed, with SDKs in Python, Go and Java and OpenAI SDK compatibility (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 6. Transparency & trust, 60 out of 100, up to 3.5 more on the total

Made of editorial 61, provenance 59.

Why it scored 60: Closed service under a public Customer Agreement and Service Specific Terms. The General Terms for AI Services, which the model terms rely on, did not load for us, so output ownership is not established (13). The data processing page and the ModelArk terms agree that customer data is not used for training without opt-in, name 180 days' retention for filter-triggered content and list processing in Malaysia, Indonesia and the EU or EEA. The Privacy Policy excludes customer data and points to a DPA we did not read (20). A deprecation process with dated batches, named replacements and a 30-day minimum (16). A sub-processor list effective 24 September 2026 with locations and 15 days' notice of changes, which has no row for ModelArk (12).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Endpoint on the vendor's domain: ark.ap-southeast.bytepluses.com is not on byteplus.com (0 of 15)
- Terms of service: published, but our reader couldn't read it (7 of 10)
- Privacy policy: published, but our reader couldn't read it (7 of 10)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 73 out of 100, up to 2.4 more on the total

Why it scored 73: Graded on the model lines. Seedream 5.0 flash (`dola-seedream-5-0-flash-260915`) is in the September 2026 release table and the deprecation page was last updated on 10 September 2026, 29 days before the check (30). The deprecation process promises at least 30 calendar days' notice and the last two batches gave about eight weeks, but the page says each version is usually retired a few months after launch, and Seedream 3.0 went in May 2026 (12). Dated release and update pages are public, with tickets and a Discord community for support, which we did not test (8). Official SDKs are current. The Python SDK reached v3.0.61 and the Go SDK v0.9.0 on 17 September 2026 (15). Both SDK repositories have CI workflows (8).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- The model list (updated 28 September 2026) says `seedream-5-0-lite` is supported in eu-west-1, while the deprecation page says the model was deactivated in that region on 10 September 2026 and the region page lists only `seed-2-0-lite` for the EU. The batch that carries the notice is dated 8 September 2026, two days earlier, against the page's own 30-day minimum. We made no call to confirm which page is right. 3 points (https://docs.byteplus.com/en/docs/modelark/model-deprecation-notice).

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: the ModelArk product page and the rest of www.byteplus.com marketing pages are drawn by script and showed no text, so any claims there (free credits, status link, enterprise SLA) went unread.
- unchecked: the General Terms for AI Services (https://docs.byteplus.com/en/docs/legal/AI-Services-terms) returned an empty page, so output ownership and usage rules for generated images are not established.
- unchecked: the Data Processing Addendum, the platform Terms of Service, the GenAI Acceptable Use Policy and the security whitepaper were not read. No clause on benchmarking or automated access was found in the Customer Agreement, and the unread Terms of Service should be checked before any probe is run.
- unchecked: the API Explorer at api.byteplus.com was not read, so whether it publishes a machine-readable description of `ImageGenerations` is unknown.
- unchecked: PyPI and Maven download figures and release dates. Release dates come from repository tags.
- No free image quota for Seedream was found. Whether new accounts get any free images, and whether sign-up needs a card, was not established.
- The model list and the deprecation page disagree on whether `seedream-5-0-lite` still runs in eu-west-1. We made no API call to settle it.
- The lead named Seedream 3.0, 4.0 and 5.0. The API reference today covers 5.0 pro, 5.0 flash, 5.0 lite, 4.5 and 4.0, and Seedream 3.0 was deactivated on 13 May 2026.
- The Customer Agreement (section 6.2) bars holding out as a reseller or model-as-a-service provider of the Services without written consent, which matters to anyone wrapping the API for others. Recorded as a fact with no deduction.
- `lastRelease` is the date of the latest SDK tags (17 September 2026). Model releases are dated by month only.

## Weaknesses

- No status page, incident history or ModelArk SLA was found in the pages read
- No OpenAPI file, llms.txt or Markdown docs for agents were found. The docs site is drawn by script
- The model list says `seedream-5-0-lite` runs in eu-west-1, while the deprecation page says it was deactivated there on 10 September 2026
- The Service Specific Terms say BytePlus Model Services are not available in the United States
- No idempotency key was found, so a retried request that succeeded is a second paid generation
- Image URLs expire 24 hours after generation, and a watermark is added unless `watermark` is set to false

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- POST to `https://ark.ap-southeast.bytepluses.com/api/v3/images/generations` with `Authorization: Bearer $ARK_API_KEY`. Keys and model activation are per region
- Set `watermark` to false if the image must not carry the AI-generated mark. The default is true
- Download `data[].url` within 24 hours, or ask for `response_format` `b64_json`
- Check each `data[].error` in a sequential batch. One moderated image does not fail the others, and only `usage.generated_images` is billed
- Stay under 500 images a minute per model and back off with jitter on 429. Layer decomposition reserves 17 against that limit per request

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.