# Qwen3-Max-Thinking

Qwen3-Max-Thinking is an AI model released by Qwen on Jan 26 2026. It has >1T parameters and a 256k context window. Tracked results include 87.4% on GPQA Diamond, 75.3% on SWE-Bench Verified and 85.9% on LiveCodeBench.

## Facts

| Field | Value |
| --- | --- |
| Model | Qwen3-Max-Thinking |
| Developer | Qwen |
| Release date | Monday, Jan 26 2026 |
| Licensing | Closed |
| Parameters | >1T |
| Context window | 256k |

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| SWE-Bench Verified | 75.3% | Lab | Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. |
| LiveCodeBench | 85.9% | Lab | Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. |
| Humanity's Last Exam (text only, no tools) | 30.2% | Lab | accuracy on the 2,158 text-only questions without tools |
| Humanity's Last Exam (with tools) | 49.8% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. |
| GPQA Diamond | 87.4% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. |
| IFBench | 70.9% | Lab | Tests whether the AI can follow detailed instructions and satisfy multiple constraints at once. Higher is better. |
| MMLU-Pro | 85.7% | Lab | accuracy on the ten-option MMLU-Pro benchmark |

## About Qwen3-Max-Thinking

Qwen3-Max-Thinking was announced on January 26, 2026 as the reasoning revision of Alibaba's Qwen3 Max tier. It could invoke retrieval or a code interpreter when needed. Alibaba also tested a higher-compute configuration that reused insights across successive reasoning attempts.

Qwen's standard launch table reported 75.3 on SWE-bench Verified. On Humanity's Last Exam, it scored 30.2 on the text-only subset without tools and 49.8 in the tool-assisted evaluation. Alibaba served the release under the dated qwen3-max-2026-01-23 identifier.

## Questions and answers

### When was Qwen3-Max-Thinking released?

Qwen3-Max-Thinking was released by Qwen on Monday, Jan 26 2026.

### Who made Qwen3-Max-Thinking?

Qwen3-Max-Thinking was built by Qwen. Alibaba's AI lab, building the Qwen family. The most prolific publisher of open-weight models of any major lab, alongside a closed Max and Plus tier sold through Alibaba Cloud.

### What benchmark scores did Qwen3-Max-Thinking get?

Qwen3-Max-Thinking reports 7 tracked benchmark scores — SWE-Bench Verified: 75.3%; LiveCodeBench: 85.9%; Humanity's Last Exam (text only, no tools): 30.2%; Humanity's Last Exam (with tools): 49.8%; GPQA Diamond: 87.4%; IFBench: 70.9%; MMLU-Pro: 85.7%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### What is the context window of Qwen3-Max-Thinking?

Qwen3-Max-Thinking has a context window of 256k. Providers may specify separate input and output token limits.

### How many parameters does Qwen3-Max-Thinking have?

Qwen3-Max-Thinking is reported at >1T parameters.

### Is Qwen3-Max-Thinking open source?

No. Qwen3-Max-Thinking is a closed model. The weights are not published — it is available only through the provider's own API, apps, or partner platforms.

### What came before and after Qwen3-Max-Thinking?

Qwen's previous tracked release was Qwen3-Max on Sep 24 2025, 124 days earlier. It was followed by Qwen3-Coder-Next on Feb 3 2026.


---

Canonical page: https://aireleasetracker.com/model/qwen/qwen3-max-thinking
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.
