all tasks / 💳 Bigger model or better harness? (DABStep 25) / deepseek-flash · mini-loop / rs_qfuzb4bc7zr3eyc9

run · rs_qfuzb4bc7zr3eyc9 · graded 2026-09-12 18:04:55graded on sitepublished · anyone who can see the task can open this

deepseek-flash · mini-loop

deepseek-flashvia mini-loopon 💳 Bigger model or better harness? (DABStep 25)

Grader output

free-form JSON · this task's own · run here

0.480grader.score12 of 25 cases passed

score: 0.48
passed: false
total cases: 25
passed cases: 12
pass threshold: 0.8

n_wrong=12n_scored=25by_category={"easy":1,"hard":0.35}n_unanswered=1cost_usd_total=0.1443latency_ms_p95=75397n_easy_correct=5n_hard_correct=7latency_ms_total=1089488cost_cases_priced=25latency_ms_median=40215n_skipped_no_gold=0cost_unanswered_usd=0.0069cost_per_correct_usd=0.012

The task's own judge and grader ran on this site over the answers the agent submitted. This task's reference answers are not published; the site holds them and never sends them to a solver. Which model answered, what it cost and who typed remain the client's word.

Run frame

declared by the client · not validated

cases: 25
duration total · client wall clock: 1122.27s
cost · client-declared: $0.144
engine: deepseek-flash
answers: via mini-loop

run configuration

declared by the client · not validated

agent: mini-loop
launcher: tp 0.0.15.dev29+gab6189d
executor: tp 0.0.15.dev29+gab6189d
framework: mini-loop
model · declared: deepseek-flash (trap.yaml)
model config: harness=mini-loop: one run_python tool, at most 10 runs, then a final answer · thinking=on, effort high (DeepSeek default) · isolation=none: filesystem not isolated · max_tokens=32000 · snippet_timeout_s=120
solution: https://github.com/Ruqii/trapstreet-solutions@0686ef3813b3/dabstep/mini-loop-deepseek-flash
task: https://github.com/trapstreet/trapstreet-tasks@e7e12e652aa2/tasks/dabstep
trap: 0.0.15.dev29+gab6189d

environment · self-reported, not validated

os: macOS 14.0
arch: arm64
kernel: Darwin 23.0.0
cpu: Apple M1 · 8 cores
memory: 8.0 GB
python: 3.13.7

skills & tools used

not reported

data coverage

identity: collected · tp · tp/0.0.15.dev29+gab6189d · 2026-09-12 18:04:52
model: collected · dabstep-solutions · trapstreet-solutions dabstep (trap.yaml, transcripts) · 2026-09-13 23:09:39
environment: collected · tp · tp/0.0.15.dev29+gab6189d · 2026-09-12 18:04:52
reproducibility: collected · tp · tp/0.0.15.dev29+gab6189d · 2026-09-12 18:04:52
skills: unsupported · tp runs a solver process; skills are a harness concept
tools: unsupported · tp does not observe the solver's tool calls
timing: collected · tp · tp/0.0.15.dev29+gab6189d · 2026-09-12 18:04:52
usage: collected · tp · tp/0.0.15.dev29+gab6189d · 2026-09-12 18:04:52 · $0.144 · reported

cases in this run

per-case results · all 25, in case_id order

casesolutiondurationcostjudge

discuss this task →