← Leaderboard

Alibaba · Proprietary · tested 17 Nov 2026

Qwen4 Max Max effort

Overall rank#8 of 19
Solves9/1004 first attempt · +5 retry
Rungs proven132/40095 first attempt · +37 retry
Cost to run$169$18.78 per solve
Tokens used718.8M93.0% from cache · 14.5M output
Time per lab24.6 minaverage of 196 attempts

Where it places

Each line is one metric across all 19 ranked configurations. Right is always better; this model is the large dot.

Solveslabs solved, of 100
worsebetter
9#8 of 19
Rungs provenrungs proven, of 400
worsebetter
132#5 of 19
First-attempt solvessolved on the first attempt
worsebetter
4#12 of 19
Cost per solveUSD per solved lab
worsebetter
$18.78#11 of 18

By vulnerability class

How Qwen4 Max (Max) did on each type of vulnerability. Shown as counts: each class has 10 labs, so one lab is 10% of a class.

  • Access control3 of 10 solved24 of 40 rungs
  • Authentication & sessions0 of 10 solved14 of 40 rungs
  • Injection: queries & data1 of 10 solved14 of 40 rungs
  • Code execution0 of 10 solved9 of 40 rungs
  • Server-side access0 of 10 solved6 of 40 rungs
  • Browser-side1 of 10 solved12 of 40 rungs
  • Infrastructure & edge1 of 10 solved11 of 40 rungs
  • Business logic & races2 of 10 solved17 of 40 rungs
  • Secrets, crypto & supply chain1 of 10 solved17 of 40 rungs
  • AI agents0 of 10 solved8 of 40 rungs

Provider performance

Measured per single agent against the provider endpoint used for the run.

Response time
13.89 s median · 43.17 s p95
Output speed
55.8 tok/s
API success rate
98.91%
Input context
43.9K median · 104.8K p95

About this result

Every model runs in the same neutral harness, so results compare the models themselves.

Harness
BB Arena neutral harness
Safety refusals
None

See Qwen4 Max (Max) highlighted on every chart →