Each call creates one task; no batch-generation operation is documented.
Fine-tuning
Not collected
No customer weight fine-tuning contract documented.
Knowledge cutoff
Not collected
Release date
Not collected
Original model release date not established by later API documentation.
Parameters
Not collected
Architecture
Not collected
Languages
Not collected
Context caching
Not collected
No prompt-cache controls, eligibility or rates documented.
Authentication
POST /open-api/v1/auth/token with Basic base64(client_id:client_secret), then use the returned accessToken as Bearer authorization for tasks.Source for Authentication (opens in a new tab)
Token expiry is not documented in this contract. Sample credentials are not used.
Poll or receive completion/failure callbacks. This is not streamed geometry.
Data retention
Not collected
No numeric asset/log retention period established. A one-hour download URL does not imply one-hour data deletion.
Task refunds
Server faults, timeouts and parameter-validation failures are refunded or not charged; FAQ identifies a 60-minute timeout threshold.Source for Task refunds (opens in a new tab)
Service model selection, not an immutable checkpoint hash.
Access
POST https://api.hitem3d.ai/open-api/v1/split/create-task using multipart/form-data and a Bearer access token.
Regions
Processing region not verified
Authentication
POST /open-api/v1/auth/token with Basic base64(client_id:client_secret), then use the returned accessToken as Bearer authorization for tasks.Source for Authentication (opens in a new tab)
Token expiry is not documented in this contract. Sample credentials are not used.
Each call creates one task; no batch-generation operation is documented.
Task refunds
Server faults, timeouts and parameter-validation failures are refunded or not charged; FAQ identifies a 60-minute timeout threshold.Source for Task refunds (opens in a new tab)
Published terms permit use of inputs/outputs to operate the service and improve models, including training. A no-training or zero-retention agreement is not established. Download-link expiry is separate from data retention.
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.
Hi3D native API · split
general
Not verified
No prompt-caching contract or discounted cached-input rate documented.
How it works
Not collected
No prompt-caching contract or discounted cached-input rate documented.
Minimum reusable input
Not collected
No prompt-caching contract or discounted cached-input rate documented.
Lifetime & refresh
Not collected
No prompt-caching contract or discounted cached-input rate documented.
How to enable it
Not collected
No prompt-caching contract or discounted cached-input rate documented.
Usage fields
Not collected
No prompt-caching contract or discounted cached-input rate documented.
Billing rules
Not collected
No prompt-caching contract or discounted cached-input rate documented.
Limits & data boundaries
Not collected
No prompt-caching contract or discounted cached-input rate documented.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 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
Qualified inference from provider-operated named model services and terms restricting reverse engineering/derivation. Reviewed official Hitem3D and Sparc3D repository trees do not expose corresponding checkpoint weights. Exact internal weight identities and per-model licenses are undisclosed; a public research paper or generated-asset license is not a weight release.
