← Leaderboard

StepFun · Open weights · tested 17 Nov 2026

Step-5 Preview Max effort

Overall rank#14 of 19
Solves7/1002 first attempt · +5 retry
Rungs proven97/40063 first attempt · +34 retry
Cost to run$222$31.73 per solve
Tokens used1.55B97.9% from cache · 22.4M output
Time per lab25.0 minaverage of 191 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
7#13 of 19
Rungs provenrungs proven, of 400
worsebetter
97#14 of 19
First-attempt solvessolved on the first attempt
worsebetter
2#17 of 19
Cost per solveUSD per solved lab
worsebetter
$31.73#13 of 18

By vulnerability class

How Step-5 Preview (Max) did on each type of vulnerability. Shown as counts: each class has 10 labs, so one lab is 10% of a class.

  • Access control2 of 10 solved20 of 40 rungs
  • Authentication & sessions4 of 10 solved21 of 40 rungs
  • Injection: queries & data0 of 10 solved4 of 40 rungs
  • Code execution0 of 10 solved7 of 40 rungs
  • Server-side access1 of 10 solved9 of 40 rungs
  • Browser-side0 of 10 solved8 of 40 rungs
  • Infrastructure & edge0 of 10 solved4 of 40 rungs
  • Business logic & races0 of 10 solved7 of 40 rungs
  • Secrets, crypto & supply chain0 of 10 solved13 of 40 rungs
  • AI agents0 of 10 solved4 of 40 rungs

Provider performance

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

Response time
5.94 s median · 26.64 s p95
Output speed
108.6 tok/s
API success rate
99.19%
Input context
52.9K median · 166.2K 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 Step-5 Preview (Max) highlighted on every chart →