Pass@N

Pass rate

Tasks solved / tasks in scope, up to 2 attempts (strict per-set denominator).

Formula: (tasks_passed_attempt_1 + tasks_passed_attempt_2_only) / task_set_size, where both numerator terms are means across the runs.

Includes unattempted tasks as failures. Scope-aware; reflects active filters (set, category, difficulty). A model benched several times scores the average of its runs, so run count does not inflate it. Final assisted solve rate with up to 2 attempts; drill-down companion to Solve AUC@2.

84.3%
Tasks pass 195.7/232
1st: 140 2nd: 55.7 Failed: 36.3
Avg cost / task

Avg cost / task

Average LLM cost of running one benchmark task once, in USD.

Formula: SUM(cost_usd) / COUNT(DISTINCT (run_id, task_id)) across all the model's results in scope.

Use to compare operating cost across models with similar pass rates. Does not account for quality. Combine with $/Pass for a cost-efficiency view.

$0.04
Latency p50

Latency p50

Median per-task wall time (LLM call + compile + test), in milliseconds.

Formula: 50th percentile of per-task duration_ms: LLM latency + compile time + test time.

Use p50 for a typical-case latency expectation. Unaffected by outlier slow tasks. Shown as N/A for batch runs: a batch request waits in the provider queue, so no per-request model timing is recorded and the only duration left would be our own compile-and-test time, which describes the harness rather than the model.

0.0s
Avg score

Avg attempt score

Mean per-attempt score on a 0–100 point scale (partial credit). Drill-down only.

Formula: Mean of attempt scores across all results rows: SUM(score) / COUNT(*) over the results table. Each attempt earns 0–100 points based on compile + test outcomes.

Drill-down companion to pass_at_n. Rewards partial credit but not directly comparable to pass rate; use for within-model analysis.

75.0 / 100
All-runs pass rate

All-runs pass rate

Fraction of tasks the model solved in every single run (strict consistency, also written pass^n).

Formula: tasks where ALL runs produced a passing result / tasks_attempted_distinct

Measures reliability under repetition. High value means the model is unlikely to regress on a re-run, important for CI and production use. Formal name in the literature: pass^n.

77.2%
$/Pass

$/Pass

Average USD cost per solved task (any-attempt pass).

Formula: SUM(cost_usd) / number of passed (run, task) cells across all runs.

Best single cost-efficiency metric. Penalises expensive models that pass few tasks and rewards cheap models with high pass rates.

$0.0419
Latency p95

Latency p95

95th-percentile per-task wall time. Captures tail latency.

Formula: 95th percentile of per-task duration_ms across all tasks in all runs.

Use p95 to understand worst-case latency. A low p95 means the model rarely stalls, relevant for automated pipelines with timeouts. Shown as N/A for batch runs: a batch request waits in the provider queue, so no per-request model timing is recorded and the only duration left would be our own compile-and-test time, which describes the harness rather than the model.

0.0s

Overview

GPT-5.6 Terra has run on 3 occasions, attempting 232 tasks with an average score of 75.0 / 100.

Settings

Generation parameters used across this model's runs. "varies" indicates the value differed between runs.

Temperature
varies
Thinking budget
varies
Avg tokens / run (input + output)
1,210,079
Consistency
75.9%

History

1
2
3
3 runs · oldest 7d ago · latest 7d ago

Cost

meanp95

Failure modes

  • AL0104 161 Syntax error, ')' expected view all →
  • AL0000 160 App generation failed view all →
  • AL0132 81 'System' does not contain a definition for 'CreateSequentialGuid' view all →
  • AL0111 48 Semicolon expected. Add a semicolon (;) to terminate the statement. view all →
  • AL0107 46 Syntax error, identifier expected. Provide a valid name (letters, digits, and underscores only). view all →
  • AL0118 39 The name 'CreateSequentialGuid' does not exist in the current context. view all →
  • AL0105 36 Syntax error, identifier expected; 'key' is a keyword view all →
  • AL0360 36 Text literal was not properly terminated. Use the character ' to terminate the literal. view all →
  • AL0198 28 Expected one of the application object keywords (table, tableextension, page, pageextension, pagecustomization, profile, profileextension, codeunit, report, reportextension, xmlport, query, controladdin, dotnet, enum, enumextension, interface, permissionset, permissionsetextension, entitlement) view all →
  • AL0122 21 Cannot implicitly convert type 'DateTime' to 'Date'. Use an explicit conversion or change the type. view all →

Shortcomings

AL concepts GPT-5.6 Terra struggles with. Click a row for description, correct pattern, and observed error codes.

Shortcomings analysis queued

Queued for analysis. This section will populate once the run is processed.

Recent runs

Runs
StartedRunModelTasksScoreCostDurationStatus
7d ago7b948839-87e… 196/23274.9 / 100$8.0857m 17scompleted
7d ago8e6da010-5db… 192/23274.0 / 100$8.1659m 32scompleted
7d agoaf415c12-e42… 199/23276.0 / 100$8.3559m 26scompleted

See all 3 runs →

Methodology

Pass rate counts unattempted tasks as failures (strict denominator). Avg score is the mean of every attempt the model produced — failed first tries that triggered a retry contribute one observation each, pulling the mean down. See the about page for the full breakdown and unit conventions.