Sources and Verification

This page explains how Try-Qwen-AI.com selects sources, verifies claims, handles conflicting evidence, and communicates uncertainty.

Core rule: a search result, AI answer, community post, or copied table is not treated as proof until the underlying source and scope have been checked.

Source Hierarchy

PrioritySource typeTypical use
1Official Qwen, QwenCloud, or Alibaba Cloud documentationFeatures, API behavior, Model IDs, limits, prices, regions, policies, and lifecycle.
2Official legal, pricing, app-store, status, and account pagesTerms, privacy, downloads, app identity, billing, availability, and support contacts.
3Official GitHub repositories, changelogs, and release notesCode behavior, version changes, examples, known issues, and release dates.
4Official model cards on Hugging Face or ModelScopeModel architecture, license, parameters, files, intended use, and open-weight availability.
5Primary standards and researchProtocol behavior, security standards, benchmark methods, and research findings.
6Hands-on tests and browser observationsInterface behavior, reproducible workflows, latency, errors, and practical limitations.
7Reputable secondary sourcesContext, comparison, or discovery when primary sources are incomplete.
8Community reportsPossible symptoms, regressions, or edge cases that require qualification.

Primary Sources

We use primary sources whenever they can answer the material question. Examples include:

  • Official Qwen product pages and policy documents.
  • QwenCloud documentation, Model Marketplace, pricing pages, and Trust Center.
  • Alibaba Cloud Model Studio documentation, product terms, regional model lists, and error references.
  • The Qwen and Qwen Code GitHub organizations.
  • Official Qwen model cards and repositories.
  • Apple App Store and Google Play listings for official app identity and version information.
  • Official provider status, security, or compliance pages.
  • Standards bodies, original papers, and official software documentation.

Secondary Sources

Secondary sources can help identify a question, competing interpretation, historical context, or user experience. They do not override a clear current primary source.

When a secondary source is the only available evidence, we identify its limitations and avoid presenting the claim as an official fact.

Community Evidence

GitHub issues, Reddit posts, forums, and social media can reveal real symptoms that official documentation does not describe. They can also be incomplete, outdated, misconfigured, provider-specific, or impossible to reproduce.

A community report is normally labeled Community-reported. We do not generalize one user’s experience to every Qwen product, region, model, account, or runtime.

Evidence Labels

LabelRequired evidence
Documentation-verifiedA current official source directly supports the claim.
Version-verifiedThe claim was checked against a named software, app, or model version.
Browser-observedThe interface or behavior was visible during a recorded visit.
Hands-on testedThe result was reproduced under documented settings.
Community-reportedA user report exists, but it is not confirmed as universal.
Editorial analysisThe statement is our interpretation or recommendation.
Not independently reproducedEvidence exists, but we did not reproduce the result.

Verification Workflow

  1. Define the exact claim. We identify whether it concerns a model, product, provider, region, plan, date, or protocol.
  2. Locate the primary source. We prefer the page or repository that owns the fact.
  3. Check the scope. We verify that the source applies to the same model, region, account type, platform, and date.
  4. Check recency. We look for updated documentation, release notes, or replacement pages.
  5. Cross-check material facts. Prices, limits, lifecycle dates, and safety claims are compared with another official record when possible.
  6. Record the verification date. Time-sensitive pages display when they were last checked.
  7. Separate fact from inference. Our analysis is labeled and does not become an official claim.
  8. Retest when practical. We reproduce workflows or errors when access and safety permit.

What a Verification Date Means

A “Last verified” date means that the material claims were reviewed against the available sources or test environment on that date. It does not guarantee that every linked page remained unchanged afterward.

Readers making a purchase, migration, compliance, or production decision should open the current official source and account console before acting.

Handling Conflicting Official Sources

Official sources can disagree because they describe different events or scopes. Common causes include:

  • Announcement date versus API availability date.
  • Open-weight release date versus hosted-model launch date.
  • Global documentation versus a regional console.
  • List price versus a temporary promotion.
  • Mainline Model ID versus a dated snapshot.
  • Consumer product behavior versus developer API behavior.
  • Documentation update date versus original release date.

When sources differ, we state the event each date or value represents. We do not silently combine incompatible scopes.

Pricing Verification

Pricing pages record the currency, billing unit, Model ID, provider, region, plan, input tier, output tier, cache rate, tool fee, promotion, and verification date where relevant.

We distinguish list prices from limited-time discounts. Calculators are estimates; the provider’s bill remains the final billing record.

Model and Limit Verification

Context windows, output limits, reasoning budgets, supported modalities, rate limits, and tool capabilities are checked for the exact Model ID. A family name alone is not sufficient when variants differ.

Open-weight model cards and hosted API pages may describe different serving configurations. We identify which one applies.

Screenshots and Browser Observations

A screenshot can establish that an interface displayed a feature or message at a specific time. It cannot prove a permanent policy, universal rollout, or backend implementation.

Screenshots are reviewed for date, platform, account context, visible Model ID, and personal data. We avoid presenting generated mockups as official screenshots.

Testing Evidence

Hands-on test claims follow the Testing Methodology. A test report should include enough information for a reader to understand the environment and important limitations.

AI-Generated Summaries

An AI answer, generated summary, or search-engine summary is not cited as the authority for a factual claim. It can help locate a source, but the underlying page must be reviewed.

Broken, Removed, or Changed Sources

When an official link is removed, we look for a current replacement, official repository history, model card, archived release record, or another primary record. We do not keep a broken citation merely because it once supported the page.

If a historical source remains relevant, the article explains that it is historical.

Source Independence

Linking to a provider does not mean the provider approves our interpretation. We may cite a source and still criticize missing documentation, unclear scope, conflicting dates, or an unsupported marketing conclusion.

Corrections

A reader who finds a source that contradicts a page can submit it through Contact. We review the source’s authority, scope, and date under the Corrections Policy.

Limitations

Some Qwen features require an account, regional access, paid plan, preview approval, or unavailable hardware. We cannot independently test every model and product surface. Where verification is incomplete, the page should say what was and was not confirmed.

Last updated: August 23, 2026.