Qwen3-Max-Preview
Qwen3-Max-Preview places in the top three on 2 benchmarks.
Qwen3-Max-Preview
Closed weights Flagship
- MCP-Universe 18.2% #29 /31
- Earth-Bench 50.2% #6 /10
- AIME 2024 65.4% #54 /84
How does it do against its rivals?
Qwen3-Max-Preview beats Grok 4 on 3 of the 8 benchmarks they share. GPT-5 does best against it, ahead on 7 of 9.
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| NOVA-63 | 61.1% | #2 of 52 | Official leaderboard | |
| Arena Expert | 1,134 | #3 of 64 | Official leaderboard | |
| LMArena (Elo) | 1,434 | #93 of 345 | Official leaderboard | |
| MGSM | 92.2% | #19 of 64 | Official leaderboard | |
| AGI-Eval Community Instruction-Tuned Language Models | 0.73 | #32 of 94 | Official leaderboard | |
| LLM Round-Trip Translation Benchmark | 8.32 | #5 of 7 | Official leaderboard | |
| Flash Fiction Stylistic Diversity | 0.18 | #21 of 23 | Official leaderboard |
Math and science
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Earth-Bench | 50.2% | #6 of 10 | Official leaderboard | |
| MCP-Universe | 18.2% | #29 of 31 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MedQA Bias Evaluation | 87.4% | #51 of 83 | Official leaderboard |
What does it cost to run?
On the 3 leaderboards that publish a price, its best run usually costs about the same as the median model's. On every one, another model comes close to its score for less.
| Benchmark | Score | Price | From cheapest to priciest | Against the median | Close to its score for less Close for less |
|---|---|---|---|---|---|
| AIME 2024 | 65.4% | $0.022 1.0× the median | 1.0× | | |
| MedQA Bias Evaluation | 87.4% | $0.0032 1.0× the median | 1.0× | | |
| MGSM | 92.2% | $0.0023 0.9× the median | 0.9× | |
Who measured it?
Independent runs cover all 12 of its benchmarks.
- An independent run only 12
Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards