Alibaba unveiled Qwen3-Max - the largest LLM at 1 trillion parameters

What is Qwen3-Max?

Qwen3-Max is a new language model from the Qwen Team (part of Alibaba), which according to the company bypasses leading AI systems such as:

* Claude Opus 4

Kimi K2

* DeepSeek-V3.1

Test results

Alibaba has tested the model in key benchmarks for LLM:

* SuperGPQA

AIME25

LiveCodeBench v6

Arena-Hard v2

LiveBench

VentureBeat reporters noted that Qwen3-Max-Preview avoids typical LLM errors, such as:

* incorrect counting of letters in words;

* errors when comparing numbers.

In addition, the model responds significantly faster than ChatGPT and other competitors.

Where can I try Qwen3-Max?

Qwen3-Max-Preview is currently available for testing through:

* 🌐 web version of Qwen Chat

* ☁️ Alibaba Cloud API

* 🔗 OpenRouter

* 💻 AnyCoder

Usage Fee

Unlike previous Qwen models, the new version is not distributed with open weights. Developers can only work with Qwen3-Max through Alibaba's paid API or distribution partners.

Current rates:

* 0-32K tokens.

* admission: \$0,861

* exit: \$3,441

32-128K tokens

* input: \$1,434

* output: \$5,735

128-252K tokens

* entry: \$2,151

* output: \$8,602

"My first, brief, purely experimental tests show that Qwen3-Max-Preview not only avoids common LLM errors, such as incorrectly counting the "R"s in "Strawberry" and incorrectly determining which number is larger, but it also responds incredibly quickly. Faster than ChatGPT"

— VentureBeat

Totals ###

Qwen3-Max is the largest LLM to date, with 1 trillion parameters. Although not yet available in open-source format, the model is already performing well and is ready for use via Alibaba's cloud services.

➡️ If you want to test the world's largest language model, try Qwen3-Max-Preview on Alibaba's official platforms.