Qwen3 Max Thinking

Jan 2026
Qwen3 Max Thinking
Compare in
on
—per task?
$1.20 / $6per 1M in / out?
—output speed?
256Kcontext

Rankingin thinking mode?

Strong at

A
Science#68 of 312Instructions#47 of 245
B
Tool calling#67 of 242Reasoning#93 of 334Long context#93 of 273
C
Knowledge#164 of 318

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperCommand A+thinkingOpen×6.7-75%$0.30 / $1.50Context: 256K → 128KBetterGPT-6 Lunalow+12%-92%$0.005Function calling only with none · Chat Completions+3OOpenMiniCPM5 1Bnon-thinking×105.1——Context: 256K → 131KCheaperGrok 4.3medium×5.6-35%$1.25 / $2.50CheaperQwen3.8 27Bnon-thinkingOpen×5.2-60%$2.49max_tokens range · QwenCloud+4

History

2 versions
9192939495Jan 26Release dateAA Omniscience · Hallucination Rate (lower is better)PreviewQwen3 Max Thinking
9192939495Jan 26Release dateAA Omniscience · Hallucination Rate (lower is better)Preview

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam28#93
AA Omniscience Index-36.6#165
CritPt (physics)1.7#108
AA Intelligence Index 4.3.2≈21.3#113
AA Omniscience · Accuracy29.8#79
AA Omniscience · Hallucination Rate94.7#300
AA-LCR (long context) 1.174.3#93
MMLU-Pro · official boardlab85.7
GPQA Diamond86.1#68
Coding
Terminal-Bench Hard (version unstated)24.2#77
Agents & Tool Use
τ²-bench · AA83.6#69
Writing & Language
IFBench (instructions)70.7#47

FAQ

How much does Qwen3 Max Thinking cost?›

$1.20 per 1M input tokens and $6 per 1M output tokens (cheapest host: DeepInfra).

Is Qwen3 Max Thinking free to use?›

There is no free API; the cheapest paid price is $1.20 per 1M input tokens.

How good is Qwen3 Max Thinking, and what is it best at?›

#93 of 334 in Chat & Knowledge, #77 of 239 in Coding, #67 of 242 in Agents & Tool Use, #47 of 245 in Writing & Language. Strongest at: Science, Instructions.

Is Qwen3 Max Thinking better than Qwen3 Max?›

On AA Omniscience · Hallucination Rate, Qwen3 Max Thinking scores 94.7 vs 89.9 — lower.

How can I use Qwen3 Max Thinking?›

Through the API of DeepInfra.

What is the context window of Qwen3 Max Thinking?›

256K tokens.

When was Qwen3 Max Thinking released? Is there a newer version?›

Released 2026-01-26. It has been superseded by Qwen3.6 Max Preview.

Is Qwen3 Max Thinking open source?›

No, Qwen3 Max Thinking is a closed model available through an API. The closest open-weights option here is MiniCPM5 1B.

Can Qwen3 Max Thinking read images?›

No, it accepts text only.

What is Qwen3 Max Thinking?›

Qwen3 Max Thinking is a closed language model from Alibaba (Qwen), released Jan 2026, the successor to Qwen3 Max.

What are the best alternatives to Qwen3 Max Thinking?›

Command A+ (cheaper), GPT-6 Luna (better), MiniCPM5 1B (open weights).