MODELLIX editorial cover reading Seedream 4.0 API with subtitle Integration Guide, Pricing and Code: an amber key at a glass API gateway with a two-line integration spec capsule

Yes, the Seedream 4.0 API still exists and is callable today — but only if you know where. ByteDance’s official international cloud, BytePlus ModelArk, still lists Seedream 4.0 at $0.03 per generated image under the model ID seedream-4-0-250828, with the full request reference still documenting it (all figures re-pulled on September 6, 2026). Meanwhile, most third-party marketplaces have quietly moved on: OpenRouter’s Seedream 4.0 page now returns 404, and Modellix — the aggregator that publishes this article — has de-listed Seedream 4.0 from its catalog and docs, as you’ll see in the availability map below. This guide gives you the honest 2026 picture: the official route that still works, the exact endpoint and code, current pricing, and a clear answer to the question none of the ranking pages address — whether a new integration should pick Seedream 4.0 at all.

Seedream 4.0 in September 2026: still sold, now the previous generation

Seedream is ByteDance’s image-generation family (Seedream 4.0 → 4.5 → 5.0). Seedream 4.0, released in September 2025 as a unified text-to-image, image-editing, and multi-image model, was the version that put ByteDance at the top of the Artificial Analysis image-editing leaderboard. The official product page is still live and still markets those capabilities: batch generation, prompt-based editing, style transfer, and multi-image fusion.

What changed is its position in the lineup. ByteDance has since shipped Seedream 4.5 (December 2025) and the Seedream 5.0 line (Lite, Pro, Edit, and Multi-Reference variants). The image-generation API reference now opens with Seedream 5.0 Pro’s new capabilities (web-search grounding, layer decomposition) and groups Seedream 4.0 at the bottom of the supported list — still supported, but clearly the previous generation. On the official route that distinction matters less than you’d think: the model page still sells it, the endpoint accepts it, and billing continues at the standard rate. On the marketplace route it matters enormously, which is the gap this guide exists to close.

Two naming traps first, because this SERP mixes them constantly. Seedream is the image family; Seedance is ByteDance’s video family — separate product, separate endpoints, per-second billing. And consumer apps (Dreamina, 即梦) are not the API; this article covers the API only.

What the Seedream 4.0 API actually is: one model, three jobs

Seedream 4.0 is a single model that does text-to-image, image-to-image, and image editing in one call surface. The official BytePlus model page documents the version and the modes:

Fact Seedream 4.0 (official, 2026-09-06)
Model ID seedream-4-0-250828 (date-suffixed — see the version-rotation warning below)
Jobs Text-to-image · single/multi-image editing · multi-image fusion (2–10 reference images on the model page; up to 14 in the shared API reference)
Output sizes 1K / 2K / 4K, or explicit pixels; default 2048×2048; range 1280×720–4096×4096
Batch sequential_image_generation: auto — up to 15 related images per request, input + output ≤ 15
Rate limit 500 images per minute
Billing Per successfully generated image — failed generations are not charged

Two integration-relevant details from the same page: the model’s rate limit is stated as 500 images per minute, and billing counts generated images, not requests — which matters for how you think about retries in a batch pipeline. ByteDance also publishes the model’s capabilities and prompt guide on its Seedream 4.0 product page; the API reference links a shared “Seedream 4.0–5.0 prompt guide” for prompt style, and recommends keeping prompts under roughly 300 Chinese characters or 600 English words.

Seedream 4.0 API pricing: the official number, pulled today

The authoritative figure is on the BytePlus ModelArk page for Seedream 4.0, read September 6, 2026: $0.03 USD per generated image, with no resolution multiplier — a 1K image and a 4K image bill the same. The page states charges are based on the actual number of generated images and that failed generations are not charged.

That flat per-image rate is the whole pricing story on the official route — there is no token meter on this model, and input prompts and reference images are not billed separately. A 10-image batch is 10 × $0.03 = $0.30; a 1,000-image product catalog is $30. The arithmetic is the easy part; the version caveats below are where budgets actually go wrong. For a deeper cost breakdown across every route and resolution rule, our separate Seedream 4.0 price analysis has the full table — this guide keeps the one number you need and focuses on getting the API working.

The official route: BytePlus ModelArk, step by step

The international path to the Seedream 4.0 API is BytePlus ModelArk (ByteDance’s cloud for non-China infrastructure). The setup has two steps people underestimate — funding and activation — so the order below matters:

  1. Create a BytePlus account and fund it. BytePlus gates model activation behind an account with a funded balance or a purchased resource pack; its own tutorials quote the balance threshold differently over time (commonly “above USD 30”), so confirm the current figure in your console rather than trusting a blog.
  2. Generate an API key under the console’s API-key management. The key is shown once — copy it into your secrets manager, never into code.
  3. Activate Seedream 4.0. Find the model in model management and activate it. The activation step refuses until step 1 is satisfied — this is the classic “my key is fine, why 401” trap.
  4. Call the endpoint. Unlike ByteDance’s video models, image generation is a single synchronous request — one POST returns the image URL.

The endpoint, from the official image-generation API reference cited above (September 6, 2026):

Region Base URL
Asia Pacific (ap-southeast-1) https://ark.ap-southeast.bytepluses.com/api/v3
Europe (eu-west-1) https://ark.eu-west.bytepluses.com/api/v3

The full request path is POST {base}/images/generations. The second regional door (eu-west-1) is new relative to what most 2025 tutorials show, and it is worth using when your workloads or data residency sit in Europe.

Working code: curl, Python, and image-to-image

curl — text-to-image. The official route accepts an Authorization bearer key and a JSON body with model, prompt, and size:

1
2
3
4
5
6
7
8
curl https://ark.ap-southeast.bytepluses.com/api/v3/images/generations \
  -H "Authorization: Bearer $ARK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-4-0-250828",
    "prompt": "A miniature yurt village at night, warm orange light in the windows, deep blue sky",
    "size": "2K"
  }'

Python — same call with requests:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
import os
import requests

resp = requests.post(
    "https://ark.ap-southeast.bytepluses.com/api/v3/images/generations",
    headers={"Authorization": f"Bearer {os.environ['ARK_API_KEY']}"},
    json={
        "model": "seedream-4-0-250828",
        "prompt": "A miniature yurt village at night, warm orange light in the windows, deep blue sky",
        "size": "2K",
    },
    timeout=120,
)
data = resp.json()["data"][0]
print(data["url"])   # expires within 24 hours — download it now, not later

Python — image-to-image / editing, passing one or more reference images base64-encoded in the image field:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
import os
import base64
import requests

with open("product_shot.png", "rb") as f:
    b64 = base64.b64encode(f.read()).decode()

resp = requests.post(
    "https://ark.ap-southeast.bytepluses.com/api/v3/images/generations",
    headers={"Authorization": f"Bearer {os.environ['ARK_API_KEY']}"},
    json={
        "model": "seedream-4-0-250828",
        "prompt": "Swap the background to a clean white studio, keep the product exactly as-is",
        "image": [f"data:image/png;base64,{b64}"],
        "size": "2K",
    },
    timeout=180,
)
print(resp.json()["data"][0]["url"])

Parameters that matter for Seedream 4.0 specifically: size accepts 1K/2K/4K or exact pixel dimensions; response_format can return url or b64_json; sequential_image_generation: auto produces a related batch in one request (input references + output ≤ 15 images); and stream: true returns each image as it finishes instead of waiting for the whole batch.

Read the model ID from configuration, never hard-code it. ByteDance version-suffixes these IDs (seedream-4-0-250828 — the suffix is a date), documents the current version via the ListFoundationModelVersions API, and when a model is updated the old ID can stop resolving or point at a different revision. This is how Seedream integrations silently break after a ByteDance release. seedream-4-0 without the suffix is not a reliable call target.

Where you can still call Seedream 4.0 — and where you can’t

The availability map below is the part every ranking page hides, because it is uncomfortable for the people selling routes. All statuses were checked live on September 6, 2026:

Route Seedream 4.0 status (2026-09-06) Notes
BytePlus ModelArk (official) ✅ Live — $0.03/image Model page and API reference both still document it
Volcengine Ark (domestic China) ✅ Live Same model family on ByteDance’s China-region cloud
DeepInfra ✅ Live — $0.04/image OpenAI-compatible endpoint, model ByteDance/Seedream-4 (API page)
OpenRouter ❌ Removed — 404 The old /bytedance-seed/seedream-4.0 page is gone; current listing is Seedream 4.5 at $0.04/image
Modellix ❌ De-listed — page state unstable ByteDance image line on Modellix now leads with Seedream 5.0; 4.0 is gone from the catalog listing and docs, though the legacy model-page URL can still intermittently serve a live page
Seedream 4.0 API access map: BytePlus open, two provider routes live, two marketplaces removed, feeding one model core

Figure: the Seedream 4.0 access map, September 2026 — official and a few providers still serve it; the big marketplaces have rotated to 4.5 and 5.0. Concept diagram generated for this article, not a screenshot.

An honest disclosure that belongs in this exact table: Modellix runs this blog and has a commercial interest in you using its API — and Seedream 4.0 is not something we promote or route new traffic to. We checked the ByteDance provider page on September 6, 2026: the Seedream image family Modellix lists is now Seedream 5.0 (Lite, Pro, Edit, and Multi-Reference variants, e.g. Seedream 5.0 Lite). The old seedream-4.0-t2i and seedream-4.0-i2i model pages are gone from that listing and their docs pages return 404; the direct legacy URLs can still intermittently render a live page with a price, so the honest one-liner is “de-listed, not deleted” — but we do not treat 4.0 as a supported offering and we are not going to list a price here for a page that may not answer. If your pipeline is locked to Seedream 4.0’s exact outputs, the official BytePlus route above is your path (or, for a legacy-route option, verify the Modellix legacy page state live the day you need it); if you are choosing a ByteDance image model fresh in 2026, the last section of this article is the one that matters — and the current generation is what Modellix exposes: Seedream 5.0 models as submit-and-poll tasks behind a single REST contract, documented in the Modellix API documentation.

Errors, limits, and the details docs pages skip

Most failed first calls to the official route are not model errors. The ones to know before you hit them:

  • 401 Unauthorized — wrong, expired, or mis-regioned key. BytePlus and Volcengine keys are separate; a key from one never authenticates on the other.
  • 403 / model not activated — the classic BytePlus trap: account and key are fine, but the model was never activated after funding.
  • Insufficient balance — per-image billing draws against the balance; the amount that cleared activation is not free budget.
  • Model ID not found — ByteDance rotated or version-suffixed the ID. Read it from configuration and re-check after each ByteDance release.
  • Moderation rejection — on the official route a rejected image is skipped without being billed (usage.generated_images counts only successes), but the request still reports a failure. In sequential mode a rejected image is skipped and the rest of the batch continues; an internal error (500) stops the batch.
  • Expired output URL — generated image URLs live 24 hours on BytePlus. Persist images when the response lands, not at the end of your pipeline.

The image-generation API reference cited above publishes the error-code table and the per-model parameter notes; the 24-hour URL expiry is stated in the same reference under the response schema. Treat all three — version rotation, activation state, and URL expiry — as production concerns, because none of them surface in a happy-path test.

Seedream 4.0 vs 4.5 vs 5.0: which API should you integrate in 2026?

The honest decision framework, based on the live lineup above (official list prices from the BytePlus billing page, read September 6, 2026 — all bill per successfully generated image, inputs free):

Version Model ID Official price Why you would pick it
Seedream 4.0 seedream-4-0-250828 $0.03/image You already generate with it and its outputs are baked into your product or your prompt engineering
Seedream 4.5 seedream-4-5-251128 $0.04/image The editing/typography-focused successor; the middle step up from 4.0
Seedream 5.0 Lite seedream-5-0-lite-260128 $0.035/image Reasoning-enhanced generation with web-search grounding — ByteDance’s current default for fresh integrations
Seedream 5.0 Pro current Pro revision Varies by generation scenario and pixel tier The flagship for professional visual-production workloads; ByteDance’s docs now lead with it
Seedream family ladder: 4.0 at the base as previous generation, 5.0 Lite and Pro on top as current default

Figure: where Seedream 4.0 sits in 2026 — still on the official price sheet, but two generations behind the models ByteDance is actively developing. Concept diagram generated for this article.

Three rules of thumb:

  1. New integration, no legacy constraint → skip 4.0. Everything ByteDance is shipping now — web-search grounding, deeper instruction following, the editing tooling — lands on 4.5 and the 5.0 line first. Starting a fresh pipeline on the previous generation buys you a migration in a few months. If you want the ByteDance image API on one key today, the 5.0 family is what aggregators (including Modellix’s Seedream 5.0 Pro page) actually serve — see our Seedream 4.5 API integration guide for the middle generation’s workflow, and the Seedream series hub for the full family inventory.
  2. Existing 4.0 pipeline, outputs you can’t afford to disturb → stay on the official route. The $0.03/image rate is still the cheapest official listing in the family, the endpoint is stable, and BytePlus has not signaled a retirement date. Just read the model ID from config and budget for the eventual move.
  3. Don’t confuse family with vendor. If your query was about generating video, that is Seedance, not Seedream — different API, different billing unit. And if you are comparing Seedream against non-ByteDance image models, our Nano Banana vs Seedream analysis does that comparison honestly.

The bottom line: “Seedream 4.0 API” in September 2026 means “the previous generation, still officially for sale at $0.03 per image, increasingly absent from third-party catalogs.” If you searched this because a 2025 tutorial told you to integrate it, the tutorial is not wrong — but the current-generation default has moved, and now you know exactly where each route stands.

Seedream 5.0 API on Modellix

See the live request schema, model IDs, and per-call pricing for Seedream 5.0 Lite and Pro on the Modellix REST API.

View Docs

Generate with Seedream 5.0 Today

Log in to Modellix to call Seedream 5.0 and 210+ other image, video, and audio models through one API key.

Login

Frequently Asked Questions

Is there a free Seedream 4.0 API?

No official free tier exists — every output image bills at $0.03 on the official route, and activation itself requires a funded BytePlus account. Some aggregators offer trial credit for new accounts; check each provider’s current terms. Input prompts and reference images are free, which is the only genuinely free part.

Does Seedream 4.0 have an API key?

Yes — you create it in the BytePlus console (API-key management) after funding your account, then send it as a Bearer token to https://ark.ap-southeast.bytepluses.com/api/v3/images/generations. Aggregator routes like DeepInfra issue their own keys for their own endpoints.

Is Seedream 4.0 on Hugging Face?

No — ByteDance publishes no open weights for Seedream 4.0, which is why “seedream 4.0 huggingface” searches find re-uploads or third-party wrappers rather than official checkpoints. It is a closed API model.

What resolution can the Seedream 4.0 API output?

1K, 2K, or 4K (or explicit pixels within 1280×720–4096×4096, default 2048×2048). Price is flat at $0.03/image regardless of resolution.

Can I use the Seedream 4.0 API from the USA?

Yes — BytePlus ModelArk is the international route (region ap-southeast-1, plus the newer eu-west-1 door) and requires no China-region account. Volcengine Ark is the domestic alternative for infrastructure inside mainland China.

What is the difference between Seedream and Seedance?

Seedream is ByteDance’s image-generation family (text-to-image and editing, flat per-image billing). Seedance is the video-generation family (per-second billing, async task endpoints). Same vendor and naming prefix, different API surface — don’t send a Seedream request shape to a Seedance endpoint.

Is Seedream 4.0 still supported?

On the official route, yes: BytePlus still documents it, sells it at $0.03/image, and lists it in the current image-generation API reference alongside 4.5 and the 5.0 line. On third-party marketplaces the answer is increasingly no — OpenRouter’s 4.0 page 404s and Modellix has de-listed the model from its catalog listing and docs (its legacy page URL may still answer intermittently) — so check the specific provider you plan to use.

Seedream model status and pricing reflect public information as of September 6, 2026 and change frequently — ByteDance has rotated model IDs and marketplace listings within a single quarter, and Modellix’s own listing of the old 4.0 pages was observed to be in flux on the same date. Re-validate against the official model and pricing pages before committing budget. This article was written by Modellix, an API aggregator with a commercial interest in ByteDance’s current-generation models — the Seedream 4.0 integration material above is vendor-neutral, and we state plainly that Seedream 4.0 is de-listed from Modellix’s catalog rather than promoted as a current offering. Access 210+ image, video, and audio models through one API key at www.modellix.ai.