BuilderBuilder
GPT-6.1 Sol · Sekiro main menu
The episode says the GPT models use an image-generation shortcut while Claude builds the scene in-engine, and that this is why the cost and time differ. Opus max is partial: 4h 14m 39.541s and $93.87 deduplicated API-equivalent, with the contestant visual review failed. A later high-effort Opus attempt is complete at 46m 30s and $11.17 and stays related, because the large cost gap matches the max run. Sol is $1.57 with verdict QUALIFIED. Sonnet is partial_user_stopped at $70.13. Supplied source capture Unaided model Open this round
Measured harness ledgerPublic result
GPT-6.1 Sol

Sekiro main menu — GPT-6.1 Sol

Recreate a high-end cinematic Sekiro-style main menu in the browser from the pinned reference and task contract.

Max reasoningHeadline result
Workflow cost
$1.57
Wall-clock
51m 06s
Processed tokens
Total input 6,052,199 (fresh input 179,431 + cache read 5,872,768 + cache creation 0). Output generation 62,360 is not included in that input total. Reasoning subset 26,347 is inside output and is not added. No processed total is recorded.
Record state
additive_harness_projection
Public summary

GPT-6.1 Sol Sekiro main menu: 51m 06s, $1.57 API-equivalent estimate, Total input 6,052,199 (fresh input 179,431 + cache read 5,872,768 + cache creation 0). Output generation 62,360 is not included in that input total. Reasoning subset 26,347 is inside output and is not added. No processed total is recorded..

Run identity and stack
  • Result ID: sekiro-main-menu-browser-gpt-6.1-sol-max-20261001
  • Technical model: gpt-6.1-sol
  • Provider: OpenAI Codex
  • Client: Codex CLI
  • Stack: Browser recreation
  • Stack: Pinned visual reference
Cost basis
  • Clock is measurements.model_invocation_wall_seconds. Cost is api_cost.total_usd. Actual subscription charge is null.
  • Source: measurements.token_usage.total_input at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Validation evidence
  • Result: QUALIFIED
  • Verification: Recorded status is session_finished_pending_acceptance. Pending, partial, stopped, qualified, and model-finished states stay unjudged. Additive projection from harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e. Broad harness sync was not run. Clock is measurements.model_invocation_wall_seconds. Cost is api_cost.total_usd. Actual subscription charge is null. Source: measurements.token_usage.total_input at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Recorded caveats
  • Verdict QUALIFIED while acceptance is pending. Not the older GPT-6 Sol menu ledger.
  • Clock is measurements.model_invocation_wall_seconds. Cost is api_cost.total_usd. Actual subscription charge is null.
  • Source: measurements.token_usage.total_input at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Visible evidence gaps
  • additive projection; broad harness sync was not run
  • source harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e
Public result only

This result keeps its public summary and evidence, but it does not currently have a matching Builder test with prompts, projects, Harness workflows or skills.

Browse BuilderReturn to GPT-6.1 Sol
RemakeBenchResearch console