Measured · Study1views

acceptance row3 1790537364808

MBPP · 3 tasks · 1 harness · 1 model

MBPP achieved the best efficiency performance in this shootout.

Ranked shootout, no baseline arm: cells ranked by the composite score — the benchmark's own graders, then efficiency — claude-code / claude-haiku-4-5-20251001 leads.

Abstract

This single-arm shootout evaluated MBPP across 3 tasks on one harness/model cell. The treatment delivered a median cost of $0.05 (mean $0.05), median tokens of 35.3k (mean 36.7k), and median duration of 26.6s (mean 35.3s). The benchmark was decided on efficiency, with claude-code / claude-haiku-4-5-20251001 emerging as the leader across all cells.

The result

Color by

claude-code

  1. 1claude-code / claude-haiku-4-5-2025100175.0best

Overall score per cell, 0–100 points (not a pass rate): 75% benchmark pass rate (the benchmark's own graders carry the outcome share — no judge grades or evals on this board) + 25% efficiency (cost · tokens · duration, vs the board's best arm). Absent components renormalize. Best at the top — the board's best setup is tagged.

Leaderboard

Cells ranked by the composite score — outcome (the lenses that graded) × efficiency — absolute values, no baseline arm. Tokens, cost and duration are per-cell medians, the efficiency the score folds.

#HarnessModelScorePass rateQualityEvalsGoalTokensCostDurationRuns
1claude-codeclaude-haiku-4-5-2025100175.067%−−100%35.3k$0.0526.6s3/3

Quality × efficiency clusters — normalized across the study

Every graded run, standardized across the study — each task ran once, so there is no field per task to compare against: → right = fewer tokens than the study's average run, ↑ up = higher quality index (85% the benchmark's own pass/fail + 15% outcome) than average. The field pools every harness and model, so a setup's left–right position largely reflects its own token habits. Dots are runs; each harness logo is a harness × model × arm centroid — hover it for the model and averages. Up-right wins.

Task
3 of 3 runs match
Show
Color by

claude-code

better · cheaperworse · pricier
← pricier than the fieldefficiency (σ, per task)cheaper than the field →

Every metric, per harness × model

Benchmark pass rate

  1. claude-code / claude-haiku-4-5-2025100167%

Task completion

  1. claude-code / claude-haiku-4-5-20251001100%

Cost per run

  1. claude-code / claude-haiku-4-5-20251001$0.05

Tokens per run

  1. claude-code / claude-haiku-4-5-2025100135.3k

Duration per run

  1. claude-code / claude-haiku-4-5-2025100126.6s

Distributions

Every completed run is one dot — the spread the averages hide. Click a dot to replay that run's journey.

Cost per run

Your setup
code/claude-haiku-4-5-20251001
0$0.07

Token composition — average per run

InputOutputCache readCache write
code/claude-haiku-4-5-20251001 · treat
36.7k

Statistics

MetricArmnMeanMedian (pooled)MinMaxStd dev
Costtreat3/3$0.05$0.05$0.05$0.07$0.01
Durationtreat3/335.3s26.6s24.4s54.8s17s
Total tokenstreat3/336.7k35.3k35.1k39.6k2.6k
Output tokenstreat3/32.4k1.1k8105.3k2.5k
Cache-read tokenstreat3/300000
Cache-write tokenstreat3/334.3k34.3k34.2k34.3k87
Turnstreat3/322220

n = runs carrying the fact / completed runs in the arm; every statistic runs over present facts only — a missing fact is never counted as 0. Std dev is the sample form (n−1), withheld below n = 2. Every figure here POOLS all runs. The typical-run figures in the abstract, the head-to-head decision and the per-setup panels take each task's median first and then the median across tasks, so a task that ran more often never outweighs one that ran once; the two medians can differ. A promise study's headline (“cut average tokens”) compares per-arm means.

What each setup did

Derived from each run's recorded tool calls — not from a model's description of the run — and aggregated per setup, so a behavior seen across several runs is stated once with its rate. Open a finding to see the runs behind it, each linked to its journey at the step where it happened.

Task
Pick a finding or a setup to list the runs behind it.
claude-code · anthropic · claude-haiku-4-5-20251001 · Your setup1 finding
  • 3 of 3 runs · 3 completed

Task text withheld

Task text withheld — this benchmark is guarded and its tasks are not republished here.

Tasks

Runs

Every run is inspectable — open one to replay the agent's journey step by step, with the analyst's read underneath.

Every run, filterable… of 3Show runs
Task
… of 3 runs match
Sort
HarnessModelTaskCompletedPassTokensCostDuration

Loading runs…

Methodology

What each metric means
Completed
The run finished and the harness returned a response. It does NOT mean the answer was correct — a completed run can score zero on quality.
Denominator: Terminal runs, excluding those killed by our own infrastructure.
Quality index
A composite used only in the quality x efficiency plot, built from whichever grader scored this study's runs — the judge's rubric, the pre-registered checks or the benchmark's own verdict — plus whether the run's own outcome was success. The map's caption names the grader and its weights.
Denominator: Graded runs — runs the study's grader scored.
Infra-excluded
A run killed by our own infrastructure. It is a missing measurement, never a loss for the arm, and is excluded from every rate denominator.
Denominator: Reported as a count beside every affected panel.
Single arm
Every task runs once per harness × model cell — a shootout with no baseline arm. The readout is absolute (quality, success, tokens, cost) and the cells are ranked into a leaderboard.
One run per task
Every task ran once per cell, in one phrasing. Run-to-run variation is therefore not measured: a single task's difference can be one lucky or unlucky attempt. Every row of the reference table treats the TASKS as the unit: its intervals resample the tasks and its p-values come from a paired test over them, so they describe how much the result depends on which tasks were drawn — not how it would change if the same tasks were run again. With few tasks no difference can reach significance (five tasks cannot go below p = 1/16).
Sample size
1 cell × 3 tasks × 1 arm = 3 runs.