Qwen3 Max vs Qwen3 Max Thinking
Comparing 2 AI models · 5 benchmarks · Alibaba
Most Affordable
Al
Qwen3 Max
$1.20/1M
Highest Intelligence
Al
Qwen3 Max Thinking
77.6% GPQA
Best for Coding
Al
Qwen3 Max
44.7 Coding Index
Price Difference
1.0x
input cost range
Composite Indices
Intelligence, Coding, Math
Standard Benchmarks
Academic and industry benchmarks
Benchmark Winners
5 tests
Al
Qwen3 Max
2
- MMLU Pro
- LiveCodeBench
Al
Qwen3 Max Thinking
3
- GPQA
- HLE
- AIME 2025
| Metric | Al Qwen3 Max | Al Qwen3 Max Thinking |
|---|---|---|
| Pricing Per 1M tokens | ||
| Input Cost | $1.20/1M | $1.20/1M |
| Output Cost | $6.00/1M | $6.00/1M |
| Blended Cost 3:1 input/output ratio | $2.40/1M | $2.40/1M |
| Specifications | ||
| Organization Model creator | Alibaba | Alibaba |
| Release Date Launch date | Sep 23, 2025 | Nov 3, 2025 |
| Performance & Speed | ||
| Throughput Output speed | 26.2 tok/s | 40.3 tok/s |
| Time to First Token (TTFT) Initial response delay | 1675ms | 1709ms |
| Latency Time to first answer token | 1675ms | 51383ms |
| Composite Indices | ||
| Intelligence Index Overall reasoning capability | 55.1 | 55.8 |
| Coding Index Programming ability | 44.7 | 36.2 |
| Math Index Mathematical reasoning | 80.7 | 82.3 |
| Standard Benchmarks | ||
| GPQA Graduate-level reasoning | 76.4% | 77.6% |
| MMLU Pro Advanced knowledge | 84.1% | 82.4% |
| HLE Hard language evaluation | 11.1% | 12.0% |
| LiveCodeBench Real-world coding tasks | 76.7% | 53.5% |
| MATH 500 Mathematical problems | — | — |
| AIME 2025 Advanced math competition | 80.7% | 82.3% |
| AIME (Original) Math olympiad problems | — | — |
| SciCode Scientific code generation | 38.3% | 38.7% |
| LCR Code review capability | 46.7% | 57.7% |
| IFBench Instruction-following | 44.1% | 53.8% |
| TAU-bench v2 Tool use & agentic tasks | 74.3% | 83.6% |
| TerminalBench Hard CLI command generation | 19.1% | 16.3% |
Key Takeaways
Qwen3 Max offers the best value at $1.20/1M, making it ideal for high-volume applications and cost-conscious projects.
Qwen3 Max Thinking leads in reasoning capabilities with a 77.6% GPQA score, excelling at complex analytical tasks and problem-solving.
Qwen3 Max achieves a 44.7 coding index, making it the top choice for software development and code generation tasks.
All models support context windows of ∞+ tokens, suitable for processing lengthy documents and maintaining extended conversations.
When to Choose Each Model
Al
Qwen3 Max
- Cost-sensitive applications
- High-volume processing
- Code generation
- Software development
Al
Qwen3 Max Thinking
- Complex reasoning tasks
- Research & analysis