Chinese model APIs on one OpenAI-compatible key. Browse models
Qwen / Alibaba logo on the SmarToken Qwen API hub
Qwen / Alibaba

qwen api

Qwen API on SmarToken: separate text work from video hype

Qwen searches mix cheap multilingual text, vision apps, and Wan-class video. This hub keeps those jobs apart, shows catalog status honestly, and links the pages you need for eval and billing.

Pick a Qwen text/multimodal ID for evaluation, and treat Wan/HappyHorse as Integrating-only until media docs say Live.

0 Live · 4 Integrating · Base URL https://thesmartoken.com/v1

How to read the Qwen family without collapsing intents

“Qwen API” is not one product. Product managers hear multilingual chat; multimodal teams hear vision; SEO teams hear Wan video. A thin page that mixes all three into one CTA is why Google and humans bounce.

On SmarToken, qwen-plus is the general text candidate for bilingual apps, structured output, and everyday coding. qwen3.7-plus is the multimodal candidate when prompts include images or longer mixed context — still verify vision parameters in docs.

Wan Text-to-Video and HappyHorse Text-to-Video appear as Integrating catalog rows. They exist so roadmap and search intent stay visible. They are not permission to fire production render jobs through Chat Completions.

Alibaba Model Studio deployment IDs are not automatically SmarToken apiModelId values. Copy IDs from this catalog or GET /v1/models. Screenshot archaeology from a mainland console is a common outage cause.

If DeepSeek Flash is already Live for your gateway, use it as the cost baseline. Promote Qwen when multilingual quality, ecosystem fit, or vision needs beat that baseline on your fixtures — not because the brand feels familiar.

Catalog rows for this vendor

Video rows stay Integrating until a media endpoint is documented as Live.

ModelAPI IDContextIn / OutStatusDetail
HappyHorse Text-to-Videovideo generationhappyhorse-t2vVideo clip$2.8 / $2.8IntegratingOpen
Qwen Plusqwen-plus128K$1.2 / $4.8IntegratingOpen
Qwen3.7 Plusqwen3.7-plus1M$0.276 / $1.101IntegratingOpen
Wan Text-to-Videovideo generationwan-t2vVideo clip$2.5 / $2.5IntegratingOpen

Text vs multimodal vs video rows

DimensionText / multimodal candidatesVideo rows (Wan / HappyHorse)
JobChat, JSON, coding, image+text evalsText-to-video demand tracking
Typical IDsqwen-plus, qwen3.7-plusWan / HappyHorse catalog IDs
Status expectationCheck Live vs Integrating per rowIntegrating until media endpoint docs say Live
Client shapeOpenAI-compatible Chat CompletionsDo not assume chat endpoint returns frames
Next docThis hub + /docs/topics/ai-video-generation-api

Workload → first pick

WorkloadFirst pickWhy
Bilingual support / localizationqwen-plusScore language mix and tone on your own tickets, not English-only benches.
Image + instruction appsqwen3.7-plusConfirm vision capability and payload limits before promising UI features.
Cheap coding assistBake-off vs deepseek-chatDeepSeek may win cost; Qwen may win language mix — measure both.
Video feature marketingVideo topic + Integrating status onlyDo not put Wan into a customer SLA until Live media docs exist.

Qwen evaluation order

  1. Step 1

    Separate text fixtures from vision fixtures

    Never judge a vision ID with text-only prompts, or a text ID with screenshot payloads.

  2. Step 2

    Compare against Live DeepSeek on the same text tasks

    Promote Qwen when it wins quality or language goals worth the rate delta.

  3. Step 3

    Park video interest on the topic page

    Subscribe your roadmap to Integrating status changes instead of hard-coding render queues.

  4. Step 4

    Pin SmarToken IDs, not Model Studio nicknames

    Config and billing logs should match this catalog’s apiModelId strings.

Billing notes for Qwen traffic

Qwen rows bill through the same prepaid USD wallet as other SmarToken routes once Live.

Integrating rows may appear in the catalog with indicative rates — do not treat indicative numbers as a production invoice until status is Live and usage lands in your wallet history.

Video cost units (seconds/frames) will not match text token math. Keep media experiments on a separate mental budget even before Live.

Use /pricing for the gateway table and vendor hubs for family-specific caveats.

Minimal request shape

Example uses qwen-plus. If the row is Integrating on your snapshot, treat the call as a staging probe only. Current example ID: qwen-plus (Integrating).

const client = new OpenAI({
  apiKey: process.env.SMTOKEN_API_KEY,
  baseURL: "https://thesmartoken.com/v1",
});

const result = await client.chat.completions.create({
  model: "qwen-plus",
  messages: [{ role: "user", content: "Return one sentence confirming the model id." }],
});

Full protocol notes live in /docs#quickstart. Confirm Live IDs with GET /v1/models before raising concurrency.

Ops notes

  • Confirm Live vs Integrating on deploy day — family fame ≠ route readiness.
  • Multimodal payloads: validate size limits and content-type assumptions in staging.
  • Do not share retry storms between text chat and future media jobs.
  • When migrating from Alibaba console IDs, keep a mapping sheet in the PR description.

Qwen-specific traps

  • Assuming Model Studio deployment IDs equal SmarToken apiModelId values.
  • Putting Wan video rows into the same alerting policy as text chat.
  • Judging Qwen only on English coding benches when traffic is bilingual support.
  • Ignoring Integrating labels because the Qwen brand is familiar.
  • Promising video SLAs from a Chat Completions client.

When to open a different hub

  • You need Live DeepSeek canaries today — /models/deepseek.
  • Long-context agents are the product — /models/kimi.
  • You only need SiliconFlow migration framing — /siliconflow-alternative.

FAQ

Is Qwen Live on SmarToken?
Check the Status column on this page and GET /v1/models. Do not infer Live from Alibaba marketing pages.
Should I use Qwen or DeepSeek for cost?
Run the same task set. DeepSeek Flash is often the first Live cost baseline; Qwen wins when multilingual or vision fit beats raw token price.
Where is Wan documented?
See /topics/ai-video-generation-api for intent and honesty rules, and the Wan row here for Integrating status.
Can I reuse DashScope SDK code as-is?
Prefer the OpenAI-compatible client against SmarToken /v1 with catalog IDs. Re-test any vendor-specific parameters.