Account- and model-specific concurrency, counting queued and processing requests. Dashboard limits are authoritative; current guide describes HTTP 400 on saturation and no standard rate-limit or Retry-After headers.Source for Request quotas (opens in a new tab)
The introductory blog advertises 20 concurrent requests after key creation. This is not assumed to be every account or model limit.
Safe request retries
Optional Idempotency-Key: 1–255 visible ASCII characters without whitespace. Same account, endpoint, body and webhook reuse the original request ID; mismatches return 422. Read status_url for current state.Source for Safe request retries (opens in a new tab)
Generation submissions only. Replays are not new generations and are not prompt caching; no expiry interval is established.
Pricing coverage
Published USD model tariffs; authenticated /estimate/{model-path} gives a configuration quote. Example credits/usd values in shared documentation are illustrative, not this model’s rate.Source for Pricing coverage (opens in a new tab)
Billing basis
Separate prepaid API balance; consumer subscriptions do not fund API use. Failed, moderated and successfully canceled queued generations are refunded.Source for Billing basis (opens in a new tab)
Generic documentation describes credits; current model pages publish USD rates. No fixed credit-to-dollar exchange is inferred.
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 15 Oct 2026
Category
Usage
Rate
Unit
Media generation
720p generated image1
$0.0938
image
Media generation
1080p generated image1
$0.1875
image
Rate conditions
Published resolution-specific API price per output image. Batch size is 1 or 4; Soul ID training, if used, is a separate request. Not a consumer-plan credit price.
API content may be used for training unless the workspace opts out; processing can take 10 business days and is prospective. Enterprise agreements can differ. Request records remain while the account is active; generated files are available for at least seven days. Zero retention requires an eligible enterprise addendum. Partner/excluded-model policies can differ.
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.
Higgsfield API
higgsfield-ai/soul/standard
Not verified
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
How it works
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
Minimum reusable input
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
Lifetime & refresh
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
How to enable it
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
Usage fields
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
Billing rules
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
Limits & data boundaries
Not collected
No prompt-cache contract or discount established. Reusing a request ID or trained identity is not inference caching.
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
Higgsfield identifies this as its own model and distributes it through its hosted products. Closed-weight classification is inferred from first-party attribution, commercial access and API terms restricting weight extraction; the reviewed sources do not establish a downloadable checkpoint or an open-weight license.
