API-token calls use a separate workspace USD balance. Only completed jobs are billed; failed or cancelled jobs are not. In-flight jobs finish if funds run out, but new requests return HTTP 402.Source for Billing basis (opens in a new tab)
Processing workflow
Poll GET /jobs/{id} until completed, failed or cancelled. Pending includes backlogged, queued, scheduled, processing, sampling and intermediate-complete. DELETE /jobs/{id} can cancel queued or processing jobs.Source for Processing workflow (opens in a new tab)
Request quotas
Concurrency, backlog and request quotas depend on plan; current guide gives no numerical universal limits. Excess concurrent submissions enter the account backlog.Source for Request quotas (opens in a new tab)
Reference URL fields accept external URLs, uploaded assets or base64 data URIs, but the schema caps each string at 1024 characters. Larger inline payloads conflict with that limit; prefer uploaded or public URLs.Source for Reference media access (opens in a new tab)
POST /styles/train model=k2; required name and urls; Style/Object/Character; optional trigger_word; max_train_steps 1–2000. No image-count limit is established by this API schema.Source for Training controls (opens in a new tab)
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 15 Oct 2026
Category
Usage
Rate
Unit
Image generation
Native text-to-image1
$0.03
image
Image generation
Native with style references1
$0.035
image
Image generation
Native with moodboard2
$0.04
image
Rate conditions
Published native configuration tariff. Image-to-image-specific billing and training fees are not established by this tier.
Moodboard price also covers combining a moodboard with style references; do not add both surcharges.
General service terms provide broad content-use licenses and make submission visibility depend on plan settings. Exact API retention, residency and a no-training commitment have not been established; enterprise terms may differ.
Up to 10 image_style_references using image_url, with strength −2 to 2, default 1. Negative strength pushes away from a reference. This differs from native url and 0–1 strength.Source for Reference guidance (opens in a new tab)
Published price covers plain generation and style-reference generation. Moodboards are accepted by the schema but their host-specific rate is not published on the reviewed page.Source for Pricing coverage (opens in a new tab)
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 15 Oct 2026
Category
Usage
Rate
Unit
Image generation
Fal text-to-image1
$0.03
image
Image generation
Fal with style references1
$0.035
image
Rate conditions
Per generated image for this host and configuration. Do not inherit the native moodboard rate.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Caching
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Krea API
image/krea/krea-2/medium
Not verified
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
How it works
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Minimum reusable input
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Lifetime & refresh
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
How to enable it
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Usage fields
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Billing rules
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Limits & data boundaries
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 15 Oct 2026
fal.ai
krea/v2/medium/text-to-image
Not verified
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
How it works
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Minimum reusable input
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Lifetime & refresh
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
How to enable it
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Usage fields
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Billing rules
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
Limits & data boundaries
Not collected
No model-specific inference-cache mechanism, TTL or discount established. Reusable styles, moodboards and job records are not prompt caches.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 15 Oct 2026
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Krea documents a hosted variant with post-training and additional style/prompt systems, alongside downloadable RAW and Turbo checkpoints. Exact hosted-checkpoint equivalence and private-weight status remain unresolved. This record captures the managed service contract and does not count as an independently proven private base model.
