This is a scannable reference for prompting Qwen in Qwen Chat (free at chat.qwen.ai, plus iOS and Android apps). The flagship is Qwen3.7-Max, an agent model with a 1M-token context window — you can paste a whole report or an entire repository and it still holds the thread. Bookmark this page and copy straight from the tables and prompt blocks below.
Qwen's strengths are hard reasoning, coding, and multilingual work, and it responds best to a real problem with rich context rather than a wall of formatting rules. The key controls are the mode (Auto, Thinking, or Fast) and the tools (Deep Research, Web Search, Code Interpreter, Web Dev, file upload). If you want fully written prompts instead of a reference, jump to our best Qwen prompts roundup.
1. The universal prompt formula
Every strong Qwen prompt follows one shape: [Role] + [Task] + [Context] + [Constraints] + [Output format]. You do not need all five every time, but the more real context you give, the better Qwen reasons — and with a 1M-token window on Max, you can afford to give a lot.
| Slot | What it sets | Example fill |
|---|---|---|
| Role | The expertise Qwen adopts | "You are a senior data engineer." |
| Task | The one action to perform | "Find why this query is slow." |
| Context | Background, inputs, real data | "Postgres 16, schema and query pasted below." |
| Constraints | Limits, must-do, avoid, length | "No new indexes, keep the rewrite readable." |
| Output format | Exact shape of the answer | "Return the rewritten query, then a 2-line why." |
You are a senior data engineer. Find why the query I paste below is slow and rewrite it. Context: Postgres 16, roughly 5M rows in the orders table, the slowdown appears at peak load. Constraints: no new indexes, keep the rewrite readable. Think step by step, then verify the rewrite returns the same rows before answering. Output the rewritten query as one block, followed by a two-line note on what was slow. Query: [PASTE QUERY]Best for: any analysis, coding, writing, or planning task where you can describe the problem and picture the ideal answer's shape.
2. Modes & tools
Pick a mode, then add tools as the task needs them. Auto handles most work, but Thinking, Deep Research, and the tools each unlock a different capability. More on the deep end in how to prompt Qwen for reasoning.
| Mode / tool | What it does | When to use it | How to trigger it |
|---|---|---|---|
| Auto | Default; Qwen decides whether to think, search, or answer fast | Everyday work when you're not sure which mode fits | Selected by default in Qwen Chat |
| Thinking | Deep reasoning; shows its full chain-of-thought before the answer | Hard problems, math, debugging, strategy | Pick Thinking in the mode selector |
| Fast | Instant reply, no reasoning tokens | Quick drafts, lookups, everyday chat where speed wins | Pick Fast in the mode selector |
| Deep Research | Agent that searches many sources, resolves conflicts, returns a cited report (usually 3,000–5,000 words) | A thorough, sourced briefing on a whole topic | Turn on the Deep Research tool |
| Web Search | Pulls live web results into a single answer | Current facts, prices, recent news; ask it to cite | Toggle Web Search on before sending |
| Code Interpreter | Runs Python for data analysis, math, charts, file processing | Crunching numbers, plotting, cleaning a dataset | Enable Code Interpreter, then attach or paste data |
| Web Dev | Generates a full website or app from a natural-language prompt | Spinning up a landing page, tool, or prototype | Turn on the Web Dev tool and describe the build |
| File upload | Reads the whole PDF, DOCX, TXT, or image you attach | Summarizing, extracting, or querying a document | Click the paperclip and attach the file |
Best for: deciding in one glance whether a prompt needs visible reasoning (Thinking), a sourced report (Deep Research), a live answer (Web Search), or code (Code Interpreter).
3. Models
Qwen Chat and the API expose several models. Pick by the job: reach for Max when the context is huge or the task is agentic, and drop to the cheaper tiers for volume.
| Model | Best for | Note |
|---|---|---|
| Qwen3.7-Max | Agent work, huge documents, whole repositories, top accuracy | Proprietary flagship (May 2026) with a 1M-token context window |
| Qwen3-Max-Thinking | Hard reasoning, math, and step-by-step problems | The earlier flagship (Jan 2026); still strong for deep reasoning |
| Qwen-Plus | Balanced everyday tasks — writing, coding, general chat | Good speed-to-quality trade-off for most work |
| Qwen-Flash | Fast, high-volume, low-cost work | The cheap tier for simple, repetitive tasks at scale |
| Open-weight Qwen3 | Self-hosting and building on the model yourself | Released on Hugging Face and GitHub under Apache 2.0 |
Best for: matching cost and context to the task — Max for the big, hard jobs, Flash for cheap volume, open-weight Qwen3 when you want to run it yourself.
4. Modifier tables
These short phrases steer Qwen's role, output shape, and tone. They stack cleanly on top of any prompt above — add one line each and keep the core the same. For ready-made structures, see the Qwen prompt templates.
Role modifiers
Open with a role to set the expertise and voice. Be specific — "senior Python engineer" beats "expert".
| Phrase | Effect |
|---|---|
| "You are a senior [LANGUAGE] engineer." | Debugging, refactoring, code review, architecture calls |
| "You are a sharp copy editor." | Tightening prose, fixing tone, cutting filler without losing meaning |
| "You are a professional translator, native in [LANGUAGE]." | Natural, idiomatic translation — a Qwen strength across 100+ languages |
| "You are a data analyst." | Reading numbers, spotting trends, building comparison tables |
| "You are a patient tutor." | Explaining a hard concept simply, step by step, with examples |
| "You are a pragmatic strategist." | Business plans, prioritization, blunt trade-off analysis |
Format modifiers
Add one line to control how Qwen returns the answer.
| Phrase | Effect |
|---|---|
| "Output in clean Markdown." | Headed sections and lists you can paste anywhere |
| "Return a table with columns X, Y, Z." | Structured comparison instead of prose |
| "Give a numbered list of steps." | An ordered checklist you can follow in sequence |
| "Output valid JSON only, no prose." | Machine-readable output for a downstream script |
| "Summarize in one paragraph." | A tight answer with no headings or bullets |
| "Cite every claim with a link." | Checkable sourcing — pair with Web Search or Deep Research |
Tone modifiers
One tone line changes the whole feel of the reply. Add it near the end so it does not get lost.
| Phrase | Effect |
|---|---|
| "Keep it plain and direct." | Cuts hedging and filler; gets to the point |
| "Be blunt about weaknesses." | Honest critique instead of reflexive agreement |
| "Explain it like I'm new to this." | Simple language, defined terms, worked examples |
| "Formal, no contractions." | Professional register for reports and docs |
| "Keep it warm and encouraging." | Supportive framing for coaching or feedback |
| "No hype words, no clichés." | Strips marketing fluff from the output |
Best for: reusing a good prompt across contexts — keep the core, swap the role, format, or tone line to change who it sounds like and what shape it returns.
5. Putting it together: example prompts
Five ready-to-run prompts that combine the formula with the modifiers above. Edit the bracketed parts and send. For the full set of written-out prompts, bookmark the best Qwen prompts roundup.
Reasoning (Thinking on)
You are a patient tutor. Solve this problem and teach me the method: [PASTE PROBLEM]. Define your variables first, then think step by step and show every step of your work. Consider one alternative method and confirm both give the same answer. Verify the result against the constraints before answering. Output a numbered worked solution, then a one-line summary of the key idea.Note: switch to Thinking mode so you can watch the chain-of-thought and catch any slip early.
Deep Research (report)
You are a research analyst. Produce a thorough report on [TOPIC] as of this month. Cover the current state, the main viewpoints, the key players, the biggest recent developments, and what is still contested. Resolve conflicts between sources and say which claim is best supported. Cite every claim with a link and prefer recent primary sources. Structure it with clear sections and end with a short bottom line.Note: turn on the Deep Research tool so Qwen searches many sources autonomously and returns a structured, cited report.
Translation (multilingual)
You are a professional translator, native in [TARGET LANGUAGE]. Translate the text below from [SOURCE LANGUAGE] to [TARGET LANGUAGE]. Keep the meaning and tone exact, use natural idiomatic phrasing rather than word-for-word, and preserve any names, numbers, and formatting. Flag anything that doesn't translate cleanly and give the two best options with a one-line note on the difference. Text: [PASTE TEXT]Note: multilingual work is a Qwen strength — 100+ languages, especially strong Chinese to and from English and other Asian languages.
Code Interpreter (data)
You are a data analyst. I've attached [FILE]. Using Python, load it, clean obvious issues (blank rows, wrong types), and answer: [YOUR QUESTION — e.g. which segment grew fastest quarter over quarter]. Show the key figures, plot the trend as a chart, and state any assumption you had to make. Output a short written summary, the chart, and the exact numbers behind it. Be blunt about anything the data can't support.Note: enable Code Interpreter so Qwen actually runs the Python on your file instead of estimating.
Coding
You are a senior [LANGUAGE] engineer. Here is my code: [PASTE CODE]. Task: find the bug causing [SYMPTOM]. Think step by step, then verify your fix against the symptom before answering. Constraints: keep it idiomatic, add no new dependencies. Output the corrected code as one block, then a two-line note on the root cause and how to prevent it.Note: paste the real code and the exact error — Qwen reasons far better with concrete inputs than with a description, and on Max you can paste the whole file.
Want the full set of written-out prompts? Bookmark the roundup at /best-qwen-prompts and keep this cheat sheet open beside it.
Frequently Asked Questions
What is the Qwen prompt formula?
Role + Task + Context + Constraints + Output format. Tell Qwen who to be, what to do, the background it needs, the limits, and the exact shape of the answer. Give it a hard problem with rich context and turn on Thinking for the difficult ones; leave it on Auto or Fast for quick drafting.
How do I switch modes in Qwen Chat?
Use the mode selector above the input box in Qwen Chat at chat.qwen.ai. Auto is the default and lets Qwen decide whether to think, search, or answer fast. Switch to Thinking for deep reasoning that shows its chain-of-thought, or Fast for instant answers with no reasoning tokens.
Which Qwen model should I pick?
Use Qwen3.7-Max for agent work, huge documents, or whole repositories — it has a 1M-token context window. Use Qwen-Plus for balanced everyday tasks, Qwen-Flash for fast, high-volume, cheap work, and open-weight Qwen3 on Hugging Face when you want to self-host under Apache 2.0.
How do I trigger Deep Research or Web Search?
In Qwen Chat, turn on the Deep Research tool for an autonomous multi-source, cited report, or the Web Search toggle for a quick real-time answer. Deep Research runs an agent that searches many sources and resolves conflicts; Web Search pulls live results into a single answer. Ask it to cite sources with links.
Is Qwen free and open-source?
Yes. Qwen Chat is free at chat.qwen.ai and in the iOS and Android apps, and the open-weight Qwen3 models are released on Hugging Face and GitHub under the Apache 2.0 license, so you can self-host. The proprietary flagships like Qwen3.7-Max run inside Qwen Chat and the API.
How do the [BRACKET] placeholders work?
Anything in square brackets is a swap-in. Replace [TOPIC], [LANGUAGE], or [PASTE DATA] with your own details before sending. The rest of each prompt is written to be copy-paste ready.