Ask drawer — ChatMaestro

ChatMaestro

v2026.08

priya.n

ResultsTerse-prompt · Pilot-20 · run of Jul 16

Results — Composite & Cost

73.9

Composite mean

$2.52

Run cost

Enrollees

312

Scored rounds

Configuration	Composite	Groundedness	Cost
GPT-5	71.2	78.4	$2.51
Terse baseline	66.9	71.3	$1.98

Ask This run · Pilot-20

Read-only · your permissions

How did GPT-5 do versus the terse baseline, and show the score trend?

GPT-5 averaged 71.2 composite across 20 enrollees — +4.3 over the terse baseline (66.9). The gain is almost all in groundedness (+7.1); coherence is flat. Cost was $2.51.

By configuration TableChart

Configuration	Composite	Groundedness	Cost
GPT-5	71.2	78.4	$2.51
Terse baseline	66.9	71.3	$1.98

Composite by self-improve pass TableChart

Read-only · checked Hide query Save prompt

select cfg, avg(score_composite) c,
sum(token_cost) cost from run_result
where run_id = %(run)s group by cfg;

Suggested for this run

Which enrollees scored lowest? Cost per composite point Compare to other Pilot-20 runs

Saved questions

For results

Lowest-scoring enrollees

BUILT-IN

Cost per composite point

APPROVED

Groundedness by nudge

YOURS · DRAFT

Model win-rate table

SHARED BY ANA

Global

Most expensive experiments

BUILT-IN

Runs finished this week

APPROVED