Qwen4 Max at launch
- Qwen
- Released November 16, 2026
- 1M context
Qwen released Qwen4 Max on November 16, 2026. As of November 22, 2026, it ranks #33 of 120 models on the LLMBenchmark Index, between Qwen3.5 397B A17B (#32) and Google Gemini 3.1 Pro Preview (#34). Qwen's own announcement reports 16 benchmark results, including Terminal-Bench 4.0 (48.5%) and Humanity's Last Exam (51.7%). It costs $1.425 per million input tokens and $4.275 per million output tokens on OpenRouter.
What Qwen reported at launch
From “Qwen4-Max: Your New Teammate Has Arrived”, August 3, 2026. The maker's own numbers, not re-run by us.
-
AA-Briefcase v1.1
1686
Elo
-
AgentArena (Net Improvement)
7.5
-
Agents' Last Exam v1.1
46.7%
#4 of 5 models with a reported result
-
APEX-Agents v1.1
72.1
#1 of 2 models with a reported result
-
AutomationBench-AA
72.5%
#2 of 5 models with a reported result
-
BioMysteryBench
85
#1 of 2 models with a reported result
-
Chartography
54.2
#9 of 16 models with a reported result
-
CodeArena
1750
Elo
#1 of 2 models with a reported result
-
DeepSWE v1.1
75.7
#4 of 32 models with a reported result
-
Humanity's Last Exam
51.7%
#10 of 39 models with a reported result
-
Healthbench Professional
60
#2 of 6 models with a reported result
-
HLAB-Vals-1.1
21.3
#1 of 2 models with a reported result
Show all 16
-
JobBench
68.5
#3 of 21 models with a reported result
-
ProgramBench
31.5
#8 of 16 models with a reported result
-
Terminal-Bench 4.0
48.5%
#7 of 21 models with a reported result
-
Video-Index
68.1
#1 of 2 models with a reported result
Independent results so far
26 scores from 26 boards that test models themselves. Ranks compare Qwen4 Max with every model on the same board.
Coding (11)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| LMArena WebDev: from an image | 1675 | #6 of 54 | xHigh | LMArena (Hugging Face dataset) |
| LMArena WebDev: React | 1785 | #7 of 101 | xHigh | LMArena (Hugging Face dataset) |
| LMArena WebDev | 1751 | #9 of 109 | xHigh | LMArena WebDev |
| LMArena WebDev: HTML | 1752 | #10 of 108 | xHigh | LMArena (Hugging Face dataset) |
| ALE-Bench | 1831.0 | #16 of 111 | xHigh | ALE-Bench (Sakana AI) |
| Terminal-Bench 4 (Vals AI) | 48.5% | #17 of 63 | Mini-SWE-agent | Vals AI boards |
| Terminal-Bench 4.0 (Artificial Analysis run) | 48.5% | #17 of 117 | — | Artificial Analysis |
| Vibe Code Bench (Vals AI) | 86.7% | #18 of 110 | OpenHands | Vals AI boards |
Show all 11
| FrontierCode 2.0 | 26.5% | #21 of 62 | — | Artificial Analysis |
|---|---|---|---|---|
| GameBench 2 | 82.4 | #23 of 105 | High | DuelLab GameBench 2 |
| LiveBench coding | 69.8% | #24 of 35 | — | LiveBench |
Agentic work (6)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark (Vals AI) | 21.3% | #6 of 40 | — | Vals AI boards |
| LiveBench browser | 80.7% | #12 of 35 | — | LiveBench |
| Tax Agent Bench (Vals AI) | 40.4% | #12 of 83 | — | Vals AI boards |
| LiveBench long agentic | 46.6% | #16 of 35 | — | LiveBench |
| LiveBench research | 84.7% | #16 of 35 | — | LiveBench |
| Legal Research Bench (Vals AI) | 41.4% | #30 of 89 | — | Vals AI boards |
General (3)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| GDPval-AA | 1809 | #5 of 118 | — | Artificial Analysis |
| LiveBench data analysis | 83.4% | #5 of 35 | — | LiveBench |
| Vals Index | 63.2% | #13 of 40 | — | Vals Index |
Reasoning (3)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| LiveBench reasoning | 92.2% | #4 of 35 | — | LiveBench |
| HLE-Diamond (Artificial Analysis run) | 44.9% | #15 of 79 | — | Artificial Analysis |
| CritPt v2, hard subset (Artificial Analysis run) | 43.0% | #23 of 62 | — | Artificial Analysis |
Intelligence (1)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| AA Intelligence Index | 42.9 | #18 of 65 | — | Artificial Analysis (via OpenRouter) |
Knowledge (1)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| AA-Omniscience | 27.7 | #17 of 65 | — | Artificial Analysis |
Web design (1)
| Benchmark | Score | Rank | Setting | Source |
|---|---|---|---|---|
| LiveBench design | 79.1% | #14 of 35 | — | LiveBench |