Qwen4-Max is live: Alibaba's new flagship, 1M context, $1.50 per million tokens
Qwen4-Max reads text, images and video, reasons and calls tools across a 1M-token window. Live on CompanyFabric the day it launched, for $1.50 in and $4.50 out.
Alibaba released Qwen4-Max, the first model of the Qwen 4 generation, and it's live on CompanyFabric today. It's Alibaba's new flagship: one model that reads text, images and video, reasons through hard problems and calls tools, all across a 1M-token context window.
The numbers
| Qwen4-Max | |
|---|---|
| Model id | qwen/qwen4-max |
| Input | $1.50 per 1M tokens |
| Output | $4.50 per 1M tokens |
| Context window | 1,000,000 tokens |
| Max output | 131,072 tokens |
| Input types | Text, images, video |
| Capabilities | Reasoning, tool calling, streaming |
For scale: a full 1M-token prompt costs $1.50. That's an entire mid-sized codebase, a year of support tickets or a stack of contracts in one request, for less than a coffee.
Call it
Same OpenAI SDK, same key, one new model string:
Where it fits
- Long-document work on a budget. At $1.50 in, Qwen4-Max is one of the cheapest ways to put a whole repository or data room into a single prompt.
- Multimodal agents. Screenshots, product photos and short videos go in the same request as text, and the model can call your tools in response.
- A second opinion. Run it next to Claude, GPT and Gemini in Council [blocked] and compare answers side by side before you commit to a model.
What we didn't add
Alibaba also published a Qwen4-Max preview. It's only available on Alibaba's Token Plan subscription and is scheduled to retire on November 25, so we left it out. You'd build on it and lose it within ten days.
Full pricing, limits and code samples are on the Qwen4-Max API page [blocked].
- Qwen
- Alibaba
- New model
- Long context
- Vision
Every model in this post is behind one key and one balance. Free starter credits on signup.
Get API KeyKeep reading
Kimi K3.1 Pro in Europe: run Moonshot's flagship on Verda GPUs in Germany
Bring your own Verda key and Kimi K3.1 Pro runs on GPUs in Germany instead of Moonshot's servers in China, at a 0% fee. Same model id, same code.
Claude Code with any model: GPT, Gemini, DeepSeek and Kimi on one key
Set two environment variables and Claude Code runs on any model in the CompanyFabric library, with prompt caching intact on Claude and a budget cap on every key.
Decision models, explained: typed answers instead of JSON you have to parse
Jev, Perplexity Decider 27B, Microsoft-Decision-1 and Gemini Decision 1 return probabilities, not prose. How they work, what they cost and which one to pick.