‹ PublicAI Index
The LLM benchmark aggregator.
Qwen3.7 Max
Alibaba
Strongest in Academic knowledge (#12 of 123), weakest in Harm refusal (#240 of 292). Above par in 18 of 25 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Gemini 3.7 Flash and ahead of GLM-5.2 and MiniMax M3.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Human preference63.8−5#36/3541/1
Human preference63.8−5#36/3541/1
Knowledge54.2−8.3#43/1971/3
Academic knowledge60.1−3.8#12/1231/1
Core abilities54.6−10.7#50/1991/4
Instruction following60.7−13.1#14/621/1
Language52.1−14.5#31/621/1
Reasoning53.3−12.1#62/1953/5
Science59.8−3.7#25/1221/1
Reasoning57.6−11.4#33/1471/3
Mathematics44.7−19.1#109/1521/2
Professional51.6−10.8#78/1941/2
Finance54.5−7.9#57/1621/1
Legal51.2−14.7#82/1731/1
Medical48.2−18.3#100/1491/1
Agents51.8−14.6#103/2982/6
Knowledge work52.8−17.9#80/2082/2
Coding44.3−21.9#132/1792/6
Agentic coding44.1−22.8#126/1732/5
Safety48.2−13.2#229/3291/3
Secure code61.9−4.3#46/2701/1
Fairness52.2−17.8#92/2921/2
Toxicity avoidance52.6−6.4#156/2681/1
Jailbreak resistance33.2−31.8#238/2681/1
Harm refusal44.9−16.6#240/2921/2
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Qwen3.7 Max, left for the other.
§ 3 · Sources
Where the numbers come from
7 publications, 30 figures. Every one links to the page it was read from.
LMArena Text 1474
GDPval-AA 1135
AA-Briefcase 917
LiveBench 1.0 73.1LiveBench 1.0 · Reasoning 83.3LiveBench 1.0 · Coding 74.2LiveBench 1.0 · Agentic Coding 43.6LiveBench 1.0 · Mathematics 85.2LiveBench 1.0 · Data Analysis 71.8LiveBench 1.0 · Language 79.7LiveBench 1.0 · Instruction Following 74
SimpleBench 70.4%
Vals · Legal Research Bench 25.48%Vals · LegalBench 84.91%Vals · Finance Agent 47.78%Vals · CorpFin 63.71%Vals · TaxEval 75.31%Vals · MedCode 38.75%Vals · MedScribe 79.4%Vals · SWE-bench Verified 68.8% (Mini-SWE-agent)Vals · Vibe Code Bench 47.67% (OpenHands)Vals · Code Migration 13.07%Vals · GPQA Diamond 90.15%Vals · MMLU Pro 89.31%
Enkrypt · Jailbreak risk 26.9%Enkrypt · Harmful content risk 8.9%Enkrypt · CBRN risk 49.8%Enkrypt · Toxicity risk 4.4%Enkrypt · Bias risk 74.4%Enkrypt · Insecure code risk 8.9%
Badge
[](https://publicai.io/model-index/m/qwen3-7-max)