Qwen 4 Max
Alibaba's new flagship and the first Qwen 4. Frontier-tier reasoning, code and analysis, reads images and video, calls tools, 1M context. Lighter filters than the US labs. Thinks before every answer, so replies take a moment. A quarter cheaper than Qwen 3.8 Max, which it replaces.
The facts
A typical message: 16K tokens in, 1.5K out. Short replies cost less. Models served via OpenRouter bill the real provider cost plus our markup, so actual charges vary. 1 credit = $0.01.
Refusal Index
How Qwen 4 Max answers the Refusal Index questions. Updated Nov 17, 2026, new questions every week.
Rank 37 of 46 (1 = fewest refusals). 30 judged answers
By category
Compare Qwen 4 Max
Same facts, side by side.
What Qwen 4 Max is best for
Alibaba's flagship Qwen. Reasoning, code and analysis, and it reads images and video.
Use it for
- Writing, reviewing and fixing code
- Multi-step reasoning: math, logic, planning
- Research and analysis of long material
- Reading screenshots, photos and scans
- Understanding video
- Whole books or codebases in one prompt
Similar models
Same lab, same tier, cheaper picks. Each one has its own page.
- MiMo V2.6 ProCheaper, same tier · ~1.09 credits per message
- DeepSeek V4 ProCheaper, same tier · ~2.17 credits per message
- Qwen 3.6 Plus UncensoredAlso by Alibaba · ~2.04 credits per message
- Qwen3 235B Also by Alibaba · ~0.45 credits per message
- Kimi K3Same tier · ~3.54 credits per message
- Grok 4.7Same tier · ~5.33 credits per message
Qwen 4 Max: questions and answers
How much does Qwen 4 Max cost?
About 4 credits ($0.04) for a typical message: 16K tokens in, 1.5K out. Per 1M tokens: $1.95 input, $5.85 output. You pay per use in credits. Free credits to start.
Is Qwen 4 Max uncensored?
Partly. Qwen 4 Max is an open model with lighter filters than mainstream models: it refuses less, but it is not fully uncensored. For zero refusals, pick a model from the uncensored tier.
Can I use Qwen 4 Max via API?
Yes. Call it as qwen-4-max on the unbottled.ai API, which speaks the OpenAI and Anthropic formats. The API needs a paid account: any plan.
How big is the context window of Qwen 4 Max?
1,000,000 tokens. That is about 1,538 pages of text in one go.
Can Qwen 4 Max read images and use tools?
Yes, it reads images. Other inputs: Video. It supports tool calling: web search, memory and code in chat, function calls on the API.
Read more
- One API key, every model. OpenAI and Anthropic formats.OpenAI chat completions, the Responses API and Anthropic Messages on one endpoint. Your SDK, Claude Code and Codex CLI on every model. What it costs, and when going direct or using OpenRouter is the better call.
- One key. Every model. Codex and Claude Code included.Codex CLI, Claude Code, OpenCode, Continue, Aider and Cline on one API key, uncensored models included. One base URL, one command per tool, the gotchas we hit.
- A ChatGPT alternative without filters. You set the level.Mainstream models keep their filters everywhere. Here they sit next to open and uncensored ones, and you pick per message. What changes, what it costs, and when ChatGPT is still the better pick.
Try Qwen 4 Max now.
Free credits to start, no card needed. The chat opens with Qwen 4 Max selected and a prompt ready. Just send.