Agent setup: If your agent doesn't auto-load skills (e.g. Claude Code),
see agent-compatibility.md once per session.

Qwen Model Selector (Advisor)

This skill operates in two modes:

1. 1. Interactive advisory — asks diagnostic questions to recommend the right model (see Diagnostic Flow).
2. Cross-skill resolution — provides a fast-path model lookup for execution skills that need a model decision

without user interaction (see Cross-Skill Model Resolution).

Do not fabricate model names — only recommend models listed in this skill.
This skill is part of qwencloud/qwencloud-ai.

Skill directory

Use this skill's reference files for data and learning. Load on demand — do not fetch external URLs unless the user
explicitly asks for latest data.

Location	Purpose
INLINECODE0	Pricing overview — model categories, billing units, and link to official pricing page
INLINECODE1

Model catalog (point-in-time snapshot) |
| references/sources.md | Official documentation URLs (manual lookup only) |
| references/agent-compatibility.md | Agent self-check: register skills in project config for agents that don't auto-load |

Security

NEVER output any API key or credential in plaintext. Always use variable references ($DASHSCOPE_API_KEY in shell,
os.environ["QWEN_API_KEY"] in Python). Any check or detection of credentials must be non-plaintext: report
only status (e.g. "set" / "not set", "valid" / "invalid"), never the value. Never display contents of .env or config
files that may contain secrets.

Coding Plan Models

Users with a Coding Plan subscription have access to a
limited set of models through their coding tools only:

Model	Context	Thinking
qwen3.5-plus	1M	Yes (budget: 81,920)
kimi-k2.5

256K | Yes (budget: 81,920) |
| glm-5 | 198K | Yes (budget: 32,768) |
| MiniMax-M2.5 | 192K | Yes (budget: 32,768) |
| qwen3-max-2026-01-23 | 256K | Yes (budget: 81,920) |
| qwen3-coder-next | 256K | No |
| qwen3-coder-plus | 1M | No |
| glm-4.7 | 198K | Yes (budget: 32,768) |

Coding Plan does not include image, video, TTS, or specialized vision models. When recommending models, note if the
user's chosen model falls outside this list and they are using a Coding Plan key (sk-sp-...). If qwencloud-ops-auth is
installed, see its references/codingplan.md for the full model list and error codes.

Diagnostic Flow

Ask the user (in order):

1. 1. Content type? — text / image / video / audio / vision
2. Primary task? — generation / understanding / coding / reasoning / translation
3. Priority? — quality vs speed vs cost
4. Input size? — short / medium / long context
5. Structured output? — JSON / function calling needed?

Cross-Skill Model Resolution

When an execution skill needs to choose a model, evaluate across three dimensions: Requirement → Scenario →
Pricing. If the user explicitly specified a model, use it as given — but still verify availability; if
restricted, warn the user and suggest an alternative.

Dimension 1 · Requirement (select)

Match task capability to the right model. Use when the user's need points to a specialized model, or when the task is
ambiguous and you need to compare capabilities.

Signal	Keywords	Model
Reasoning	"think step by step", "reason", "analyze"	qwq-plus (text) · qvq-max (vision)
Coding

"write code", "implement", "debug" | qwen3-coder-plus |
| OCR / document | "extract text", "OCR", "scan" | qwen-vl-ocr |
| Long context | "long document", "large file" | qwen3.5-plus (1M context) |
| Multimodal (text+image+video) | "analyze image", "understand video" + text | qwen3.5-plus (unified multimodal) |
| Voice interaction / omni | "voice chat", "speak", "listen" | qwen3-omni-flash |
| Built-in tools | "search the web", "run code", "use tools" | qwen3-max (web search, code interpreter) |
| Image editing / style transfer | "edit image", "style transfer", "reference image" | wan2.6-image (preferred) · wan2.5-i2i-preview |
| Image-to-image fusion | "place object", "combine images", "fuse images" | wan2.6-image · wan2.5-i2i-preview |
| Style TTS | "emotion", "tone", "pace" | qwen3-tts-instruct-flash |
| Ambiguous | task doesn't clearly map to one model | compare Recommendation Matrix; ask user to clarify if needed |

Dimension 2 · Scenario (tune)

Adjust model tier based on how the model will be used.

Pattern	Signals	Guidance
Interactive / real-time	"chat", "real-time", "interactive"	Prefer flash/turbo variants; enable streaming
Batch / offline

"batch", "offline", "background" | Quality model + Batch API (50% off) |
| One-off trial | "try", "test", "experiment" | Quality model; check if free quota is still available in user's console |
| High-volume production | "production", "at scale", "high volume" | Cost-optimize: flash/turbo + context cache |
| Repeated context | "template", "same prompt", "repeated" | Enable context caching for input token discount |

Dimension 3 · Pricing (optimize)

Given the candidates from dimensions 1–2, compare costs and apply modifiers.

• - Pricing reference: pricing.md. For the latest rates, check

the official pricing page.

• - Free quota: Some models offer a limited free quota after activation. However, quotas may have been consumed,

expired, or changed. Never assume remaining free quota — always present the paid unit price.

• - Batch API: 50% off both input and output tokens for non-realtime workloads.
• Context cache: Input token discount for repeated/templated contexts.
• Tiered pricing: Some models charge more per token as input length increases — check pricing tables for

breakpoints.

• - When cost is the user's primary concern, explicitly recommend the cheapest viable model and cite the price.

Default

No signals detected, clear task → use the Canonical Default for the domain.

Domain	Default	Quality	Speed	Cost
text.chat	qwen3.5-plus	qwen3-max	qwen3.5-flash	qwen-turbo
vision.analyze

qwen3-vl-plus | qwen3-vl-plus | qwen3-vl-flash | qwen3-vl-flash |
| omni (voice+vision) | qwen3-omni-flash | qwen3-omni-flash | qwen3-omni-flash | — |
| image.generate | wan2.6-t2i | wan2.6-t2i | wan2.2-t2i-flash | wan2.2-t2i-flash |
| image.edit | wan2.6-image | wan2.6-image | wan2.5-i2i-preview | wan2.5-i2i-preview |
| video.t2v | wan2.6-t2v | wan2.6-t2v | — | — |
| video.i2v | wan2.6-i2v-flash | wan2.6-i2v | wan2.6-i2v-flash | — |
| audio.tts | qwen3-tts-flash | — | qwen3-tts-flash | — |

Degradation: If this skill is not loaded or not available, each execution skill falls back to its own built-in
default. This protocol is purely additive — it enhances model selection but never blocks execution.

Model Recommendation Matrix

Text Models

Use Case	Recommended	Why
General chat/assistant	qwen3.5-plus	Best balance of quality, speed, cost. Also accepts image/video input (multimodal). Thinking enabled by default.
Fast responses, low cost

qwen3.5-flash | 3x faster, 70% cheaper than Plus. Thinking enabled by default. | | Highest quality | qwen3-max | Strongest reasoning. Built-in tools (web search, code interpreter). Supports thinking mode. | | Code generation | qwen3-coder-next | Best balance of code quality, speed, cost. Agentic coding. qwen3-coder-plus for highest quality. | | Complex reasoning | qwq-plus | Chain-of-thought reasoning specialist | | Long documents | qwen3.5-plus | Up to 1M context. For >1M needs, see model-list.md. | | Budget/high volume | qwen-turbo | Cheapest per-token cost |

Image Models

Use Case	Recommended	Why
Best quality text-to-image	wan2.6-t2i	Latest model, sync support
Image editing / style transfer (1–4 refs)

wan2.6-image | Multi-image composition, subject consistency, 2K output, interleaved text-image | | Image editing / multi-image fusion (1–3 refs) | wan2.5-i2i-preview | Simpler prompt-based editing, subject consistency, multi-image fusion | | Interleaved text-image output (tutorials) | wan2.6-image | Mixed text+image generation | | Fast iteration | wan2.2-t2i-flash | 50% faster generation | | Flexible resolution | wan2.5-t2i-preview | Custom aspect ratios |

Video Models

Use Case	Recommended	Why
Quick video creation	wan2.6-i2v-flash	Fast, multi-shot narrative
High quality

wan2.6-i2v | Best visual quality | | With audio | wan2.5-i2v-preview | Auto-dubbing support |

Audio Models

Use Case	Recommended	Why
Highest quality	INLINECODE10	Best naturalness, emotional expression, professional scenarios
High quality + speed

cosyvoice-v3-flash | Good balance of quality and performance | | Standard TTS | qwen3-tts-flash | Fast, reliable, multi-language, cost-effective | | Controlled style | qwen3-tts-instruct-flash | Instruction-guided voice style (tone/emotion) |

Vision Models

Use Case	Recommended	Why
Best accuracy	qwen3-vl-plus	Highest vision understanding. Thinking mode supported. 256K context.
Fast analysis

qwen3-vl-flash | Quick image understanding. Thinking mode supported. | | Unified text+vision | qwen3.5-plus | Multimodal (text + image + video). Surpasses qwen3-vl series on many benchmarks. Use when both text quality and vision matter. |

Omni Models

Use Case	Recommended	Why
Voice + vision chat	qwen3-omni-flash	Text/image/audio/video → text or speech. 49 voices, 10 languages. Thinking supported.
Real-time voice

qwen3-omni-flash-realtime | Streaming audio input + built-in VAD. 49 voices. |

Pricing Guidance

• - Default pricing: pricing.md — International, USD.

For the latest rates, check the official pricing page.

• - Latest prices: When the user explicitly asks for exact/latest pricing, see sources.md for

official URLs.

• - Cost formula: Cost = Tokens ÷ 1,000,000 × Unit price. 1K Chinese chars ≈ 1,200-1,500 tokens.
• Free quota: Some models offer a limited free quota after activation — but quotas may have been consumed, expired,

or changed without notice. Always present the paid unit price first. Mention free quota only as something the user should verify in their QwenCloud console.

• - Cost tips:

- Use Batch calling for 50% off in non-realtime scenarios - Enable context cache for repeated contexts - Use flash/turbo series for non-critical tasks

Cost Estimation Disclaimer (MANDATORY)

🚨 CRITICAL — NO EXCEPTIONS: NEVER fabricate, invent, or guess any price figure. If you do not have a
confirmed price from references/pricing.md or the official pricing page, you MUST NOT output any number.
Instead, direct the user to
the official pricing page.
Outputting a made-up price is a critical failure — worse than saying "I don't know."

When responding to any cost-related query — including but not limited to price evaluation, usage estimation, budget
forecasting, or cost comparison — you MUST append a professional disclaimer. This applies regardless of language or
response format.

Required disclaimer (Chinese response):

⚠️ 费用说明：以上费用为基于官方公示单价的预估价格，仅供参考。实际费用受 Token
消耗量、上下文长度阶梯定价、Batch/缓存折扣及计费策略调整等因素影响，请以QwenCloud控制台的实际账单为准。部分模型可能提供限时免费额度，但免费额度的可用性、额度量及有效期随时可能调整，请在控制台确认您的账户是否仍有剩余额度，切勿假设本次调用免费。最新定价详见 模型定价页。

Required disclaimer (English response):

⚠️ Pricing Notice: The cost figures above are estimates calculated from officially published unit prices and
are provided for reference only. Actual charges depend on token consumption, tiered context-length pricing,
Batch/cache discounts, and billing policy updates. Some models may offer a time-limited free quota, but
quota availability, amounts, and validity periods are subject to change — do not assume this call is free. Please
verify your remaining quota in
the QwenCloud console and refer to the actual
bill for definitive costs. See Model Pricing for
the latest rates.

Rules:

• - The disclaimer must appear at the end of every cost-related response, clearly separated from the main content.
• When the estimate involves assumptions (e.g., average tokens per character, assumed context length tier), explicitly

state each assumption used in the calculation.

• - Never present estimated costs as exact or guaranteed amounts. Use hedging language such as "approximately", "estimated

at", "roughly" (or Chinese equivalents "约", "预估", "约合") throughout the cost breakdown.

• - Never tell the user a call will be free or cost $0/¥0. Even if a free quota exists, the user may have already

consumed it. Always present the paid price and note that a free quota may apply — subject to the user verifying in their console.

• - If pricing data is unavailable or uncertain, say so explicitly and link to the official pricing page. Never fill

the gap with a guess.

Available Models

All standard text, vision, image, video, audio, and coding models are available. Some models offer free
quota (verify in console).

• - Text: qwen3-max, qwen3.5-plus, qwen3.5-flash, qwen-turbo, qwq-plus, qwen3-coder-next/plus/flash, qwen-plus-character, qwen-plus-character-ja, qwen-flash-character
• Vision: qwen3-vl-plus, qwen3-vl-flash, qvq-max, qwen-vl-ocr, qwen-vl-max, qwen-vl-plus
• Omni: qwen3-omni-flash (+ realtime), qwen-omni-turbo (+ realtime)
• Image generation (text-to-image): wan2.6-t2i, wan2.5-t2i-preview, wan2.2-t2i-flash, z-image-turbo
• Image editing (requires reference images): wan2.6-image, wan2.5-i2i-preview
• Video generation: wan2.6 series (t2v, i2v, i2v-flash, r2v, r2v-flash), wan2.5/2.2 series, vace
• TTS: qwen3-tts-flash, qwen3-tts-instruct-flash, cosyvoice-v3 series
• ASR: qwen3-asr-flash, fun-asr
• Embedding/Rerank: text-embedding-v4, qwen3-rerank
• Translation: qwen-mt-plus/flash/lite/turbo

⚠️ Important: The model list above is a point-in-time snapshot and may be outdated. Model availability
changes frequently. Always check the official model list
for the authoritative, up-to-date catalog before making model decisions.
See model-list.md for a more detailed local reference.

Thinking Mode

Several models support hybrid thinking/non-thinking modes:

Model	Thinking Default	Notes
qwen3.5-plus	On	Thinking enabled by default. Use enable_thinking: false to disable.
qwen3.5-flash

On | Thinking enabled by default. |
| qwen3-max | Off | Use enable_thinking: true for complex reasoning. Built-in tools available in thinking mode. |
| qwen-plus / qwen-flash / qwen-turbo | Off | Hybrid; enable for deeper reasoning at higher output cost. |
| qwen3-vl-plus / qwen3-vl-flash | Off | Vision + thinking for complex visual analysis. |
| qwen3-omni-flash | Off | Thinking supported; audio output not available in thinking mode. |
| qwq-plus / qvq-max | Always on | Pure reasoning models; CoT always active. |

Guidance: Do not enable thinking by default for simple or conversational tasks — it increases latency and output
token cost. Enable only when the user explicitly asks for deep reasoning or the task requires multi-step analysis.

Anti-Patterns

• - Only recommend models listed in this skill — never fabricate model names.
• When unsure, use qwen3.5-plus as a safe default for text tasks.
• 🚨 NEVER invent or guess any price figure — only use pricing from references/pricing.md or the

official pricing page. If the data is not available, say so and link to the official page. Fabricating a price is a critical failure.

• - Always cite data source when providing pricing info.
• Default currency is USD. Check

the official pricing page for the latest rates. Writing in Chinese does NOT imply CNY.

• - All cost estimates must include the mandatory disclaimer — use hedging language ("approximately", "estimated").

Omitting the disclaimer is a critical violation.

• - Never assume free quota is available — free quotas may have been consumed, expired, or removed. Never tell the

user a call will cost $0/¥0. Always present the paid unit price first; mention free quota only as a possibility the user should verify in their console.

• - Do NOT proactively fetch URLs or trigger web searches — only access online sources when the user explicitly asks

for "latest" data AND reference files cannot answer.

• - Never output API keys in plaintext — see Security section above.

Online Information Sources

Reference files in this skill are point-in-time snapshots. For official documentation URLs,
see sources.md.

Do NOT proactively fetch or search online. Only access online sources when ALL of these conditions are met:

1. 1. The user explicitly asks for "latest", "real-time", or "current" data.
2. The reference files in this skill cannot answer the question.
3. The user confirms they want an online lookup.

Update Check (MANDATORY Post-Execution)

When the user asks to check for updates (e.g. "check for updates", "check version", "is there a new version", "update
skills"):

1. 1. Find qwencloud-update-check: Look for qwencloud-update-check/SKILL.md in sibling skill directories.
2. If found — run: python3 <qwencloud-update-check-dir>/scripts/check_update.py --print-response and report the result.

Use --force if the user asks to force-check.

1. 3. If not found — you MUST present the following options and wait for the user's response:

The update-check skill is not installed. It keeps you informed about new qwencloud/qwencloud-ai features and bug fixes.
1) Install now — I'll run the install command for you
2) Skip for now — do nothing this time
3) Never remind me — permanently suppress this prompt

After the user responds:

• - User chooses 1 → run:

npx skills add QwenCloud/qwencloud-ai --skill qwencloud-update-check -y, then run: python3 <this-skill-scripts-dir>/gossamer.py --dismiss

• - User chooses 2 → run: python3 <this-skill-scripts-dir>/gossamer.py --dismiss, then continue normally
• User chooses 3 → run: python3 <this-skill-scripts-dir>/gossamer.py --never-install, then continue normally

References

• - pricing.md — Pricing overview: model categories, billing units, and link to official pricing

page

• - model-list.md — Model catalog (2026-03 snapshot;

check official model list for latest)

• - sources.md — Official documentation URLs (for manual lookup only)