/ Alibaba Cloud / Tongyi Embedding Vision Flash · March 2026 Alibaba Cloud / Tongyi Embedding Vision
Tongyi Embedding Vision Flash · March 2026 The March Flash vision embedding release supports fused content, smaller vectors and configurable video sampling.
available Closed weights Embeddings
Input token limit 1K 1,024 tokens
Vector dimensions 768, 512, 256, 128, 64 See specification notes below
Input billing See rates Alibaba Model Studio · China (Beijing)
Output billing See rates Published units and conditions below
Specifications Access & pricing Caching Evaluation Sources & lifecycle
Specifications Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.
Context / input limit Not collected
Maximum output Not collected
Reasoning Not collected
Streaming Not collected
Knowledge cutoff Not collected
Parameters Not collected Languages View all 33 entries Chinese Japanese Korean Indonesian Vietnamese Thai English French German Russian Portuguese Spanish Italian Swedish Danish Czech Norwegian Dutch Finnish Turkish Polish Swahili Romanian Serbian Greek Kazakh Uzbek Cebuano Arabic Urdu Persian Hindi / Devanagari Hebrew Source for Languages (opens in a new tab)
Access & pricing Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
API model ID tongyi-embedding-vision-flash-2026-03-06
Access https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/embeddings/multimodal-embedding/multimodal-embedding. Use a Beijing workspace API key.
Regions China (Beijing) Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 14 Oct 2026 Category Usage Rate Unit Embeddings Image/video input1 CN¥0.15 1M tokens Embeddings Text input1 CN¥0.15 1M tokens
Rate conditions Published domestic-documentation CNY tariff, not a currency conversion from USD. Beijing workspace. No generated-answer estimate is inferred. Pricing source (opens in a new tab) Data & enterprise terms Account-specific regional retention and enterprise policy require confirmation.
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 INPUT Process & retain The system processes eligible input and retains reusable state.
02 / LATER REQUEST Check available state Reuse depends on matching input, session state and the cache policy.
03 / CACHE HIT Reuse & generate Reusing state can reduce repeated work. Any charges follow the billing rules below.
Alibaba Model Studio · China (Beijing) tongyi-embedding-vision-flash-2026-03-06Not supported Support evidence (opens in a new tab) 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 Classification inference: the provider names this native API release separately from public Qwen3/GTE/GME/GVE checkpoints, and no exact release was found in the reviewed official Qwen or Alibaba-NLP weight inventories. Absence is supporting evidence, not proof of private checkpoint lineage. Base-model parameter counts are not transferred to this service.
Classification evidence (opens in a new tab) Lifecycle dates Sources reviewed Model notes Reviewed language, retrieval, decision, speech, music, video and world-model releases, with regional prices and limits. Coverage is expanding; public-weight counterparts and unresolved API identities remain under review.
This pinned March release differs from the rolling selector in dimensions, fusion, resolution controls, language coverage and media limits. It is not treated as an alias.
Fusion uses a combined content object. The generic enable_fusion parameter belongs to Qwen3-VL and is not the control for these snapshots.
The model table lists configurable dimensions for this snapshot, while generic dimension prose describes the older rolling selector. Snapshot-specific rows are retained.
Check against your workload Try representative inputs, measure task success and total cost, then check the exact endpoint’s data policy.
Read our methodology
Also from Alibaba Cloud Qwen3.7-Max Qwen3.7-Max June 2026 Qwen3.7-Max Preview Qwen2.5-Max Qwen3-VL-Plus HappyHorse 1.0 Text-to-Video HappyHorse 1.0 Image-to-Video HappyHorse 1.0 Reference-to-Video HappyHorse 1.0 Video Edit HappyHorse 1.1 Text-to-Video HappyHorse 1.1 Image-to-Video HappyHorse 1.1 Reference-to-Video 