Opus 5.5 scored 68.26. GPT-6 Sol scored 45.55.

One owner-scored pilot of each model, harness and tool setup. This is not a universal model ranking or a statistical claim.

one owner-scored pilot. not a universal model ranking.

read the method download the public result (json)watch LAB 0006

BuilderBench pilot result: Opus 5.5 scored 68.26 and GPT-6 Sol scored 45.55, with API-equivalent estimate caveats. — one owner-scored pilot. costs are estimates, not bills. Sol usage is partly recorded. methods and coverage differ.

overall score

quality 70% · autonomy 20% · proactivity 10%

Claude harness

Opus 5.5

68.26/100

quality: 61.8
autonomy: 80
proactivity: 90

Codex harness

GPT-6 Sol

45.55/100

quality: 46.5
autonomy: 40
proactivity: 50

task scores

where each setup gained and lost ground

Writing · weight 20%
Opus 5.549
GPT-6 Sol30
Video editing · weight 15%
Opus 5.540
GPT-6 Sol10
Frontend design · weight 15%
Opus 5.560
GPT-6 Sol40
Motion graphics · weight 10%
Opus 5.560
GPT-6 Sol30
Browser & computer · weight 15%
Opus 5.570
GPT-6 Sol90
Research & decisions · weight 10%
Opus 5.570
GPT-6 Sol55
Intake automation · weight 10%
Opus 5.595
GPT-6 Sol90
Founder conversation · weight 5%
Opus 5.580
GPT-6 Sol40

see the task table

Task scores and weights for each setup
task	weight	Opus 5.5 Claude	GPT-6 Sol Codex
Writing	20%	49	30
Video editing	15%	40	10
Frontend design	15%	60	40
Motion graphics	10%	60	30
Browser & computer	15%	70	90
Research & decisions	10%	70	55
Intake automation	10%	95	90
Founder conversation	5%	80	40

correction

the correction is a receipt

2026-09-24 15:13 UTC · fixed before this result went public

what happened: after the scores were locked, live-chat deliverables were still flagged as missing.

why: stale missing-file flags from the offline window. the frozen files and their hashes were checked again, and they were there.

what did not change: my ratings of the work did not change. only the missing-file flags did.

Scores before and after the correction
setup	score	before	after
Opus 5.5 · Claude	quality	57.80	61.80
Opus 5.5 · Claude	overall	65.46	68.26
GPT-6 Sol · Codex	quality	44.50	46.50
GPT-6 Sol · Codex	overall	44.15	45.55

frozen setup

what was tested

pack version: 1.0.0
clock: 4 hours
reasoning: Extra High (xhigh)
speed: Standard
billing: Existing subscriptions
owner scores: 1 per setup

exact model and harness version strings were not kept in the public pilot export, so they stay unknown here.

cost

estimates are not bills

Opus 5.5

US$49.67

Provider-reported API estimates for the recorded task rows.

GPT-6 Sol

US$12.75

Partial recorded usage only; unknown attempts and shared overhead are excluded.

API-equivalent estimates are not bills. Actual incremental spend is unknown. The methods and coverage differ, so these figures do not support an exact cost ratio.

limits

what this pilot does not prove

This was one pilot scored by one owner, not a universal ranking.
The score belongs to each full setup: model, harness, tools and settings.
The two cost estimates use different methods and coverage.
Actual incremental spend remains unknown because the run used existing subscriptions.
Complete normal-speed watched-and-listened review of both full video outputs was not established.

The owner scored quality, autonomy and proactivity after a blind review. Quality is the weighted score across eight task families. The score lock happened before identities were revealed.

p.s.

get the next BuilderBench result first.

the exact prompt i typed
what came back
where it fell over

32,545 on Threads
3,752 readers
1.78K on YouTube

live count, 29 sep 2026.