Home / Inference
ModelScope API-Inference
www.modelscope.ai/docs/model-service/API-Inference/limitsInference API, freeinference/modelscope-api-inferenceJSONcomparereport an errorprovider JSON
- Free allowance
- Free API access to selected open-source LLM, multimodal and text-to-image models, described as a non-commercial, non-profit product.
- Verdict
- fails: identitythe identity required is not reachable without it
- Cost
- free tier
- Identity
- national-idno card, phone required
- Binding limit
- request-rateconcurrency per model is limited dynamically by platform load, with normal single-concurrency use as the guiding principle; not for high-concurrency or SLA workloads
- Evidence
- documented12 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-12-04checked 2026-11-03 (live); stale on 2026-12-04
| Free tier | Free API access to selected open-source LLM, multimodal and text-to-image models, described as a non-commercial, non-profit product. Quota and concurrency are adjusted dynamically with platform load. Calls can be redeemed with Magicubes: about 0.5 per call for lightweight models, about 1 for standard and about 2 for flagship. A per-day call count is not stated on the page read unit requests; metered on request-count; perpetual-free-allowance |
|---|
| Price | none stated |
|---|
| Limits | - request-rate (binding) concurrency per model is limited dynamically by platform load, with normal single-concurrency use as the guiding principle; not for high-concurrency or SLA workloads source
- volume-quota 2000 API-Inference calls per day per registered user, as announced by the operator; the current limits page gives no daily figure and says quotas adjust dynamically with platform load source
- volume-quota per-model cap reported as 200 calls per day (100 for the largest DeepSeek models), within the daily total; the operator's current limits page gives no per-model figure and says limits adjust dynamically source
|
|---|
| Binding | request-rate dynamic per-model concurrency limits; a daily total of 2000 calls is announced but absent from the current limits page |
|---|
| Identity | national-id the ModelScope account must first bind to an Alibaba Cloud account that has completed real-name verification before API-Inference works normally. A ModelScope maintainer stated the binding is open only to Alibaba Cloud China-site accounts, not international-site accounts; a China-site account registers with a mainland China mobile number and individual verification needs an identity-verified personal Alipay account. The international site verifies by passport or driver's license, but its accounts are a separate system and are not accepted for the binding. unverified: whether the China-site Alipay verification accepts a non-mainland document |
|---|
| Egress | not published |
|---|
| Interface | rest |
|---|
| Tooling | part of the text of Free tier, above |
|---|
| Composes with | free open-model inference for experiments; the per-call Magicube cost is shown next to the provider in each model's code example |
|---|
| Reliability | older models are removed from API-Inference as new models are released |
|---|
| Parsed numbers | - has-free-tier 1
- volume-quota 200 calls / day
|
|---|
Evidence 12
Unknown values 3
- egress metered: egress metering was not published
- per object cap stated: the field does not apply to this kind of offer
- retention stated: the field does not apply to this kind of offer