BuilderBuilder
GPT-6 Astra · Turbofan CAD
The episode says the Claude runs take about 44 minutes and about $13 and $14, Sol takes almost an hour and $3 to $6, and Astra costs the most. Those figures match the CAD-001 artifact metrics: Sonnet list-rate $13.28 and 44m 56s, Opus list-rate $14.07 and 44m 60s-scale 2699.6 seconds, Sol's API-equivalent range $3.24 to $6.14 and 57m 17s, and the nested unaided Astra estimate $31.78 and 57m 57s. model_finished is not recorded as a motion-contract pass. Supplied source capture Unaided model Open this round
Campaign evidence recordPublic result
GPT-6 Astra

Turbofan CAD — GPT-6 Astra

Create the CAD-001 v1.0.1 turbofan submission in the native CAD workflow.

Max reasoningHeadline result
Workflow cost
$31.78
Wall-clock
57m 57s
Processed tokens
Input including cache 21,660,559 (uncached input 709,135 + cache read 20,951,424). Output including reasoning 74,670. Reasoning subset 35,097 is inside that output and is not added. The Jev-plus-Astra token object is not included.
Record state
campaign_evidence_not_harness_ledger
Public summary

GPT-6 Astra Turbofan CAD: 57m 57s, $31.78 API-equivalent estimate, Input including cache 21,660,559 (uncached input 709,135 + cache read 20,951,424). Output including reasoning 74,670. Reasoning subset 35,097 is inside that output and is not added. The Jev-plus-Astra token object is not included..

Run identity and stack
  • Result ID: cad-001-gpt-6-astra-max-2026-09-20
  • Technical model: gpt-6-astra
  • Provider: OpenAI Codex
  • Client: Codex CLI
  • Stack: CAD-001 turbofan fixture
  • Stack: Native CAD delivery
Cost basis
  • Clock is runs.gpt_6_astra_max.wall_clock_seconds. Cost is published_rate_estimate_usd. actual_charged_cost_usd is null. No document URL is published.
  • Source: runs.gpt_6_astra_max.tokens at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Validation evidence
  • Result: nested_metrics_unaided_run
  • Verification: Recorded status is nested_metrics_unaided_run. Pending, partial, stopped, qualified, and model-finished states stay unjudged. Campaign evidence. No top-level harness ledger. Sourced metrics stay attributed in cost, time, and the caveat. Clock is runs.gpt_6_astra_max.wall_clock_seconds. Cost is published_rate_estimate_usd. actual_charged_cost_usd is null. No document URL is published. Source: runs.gpt_6_astra_max.tokens at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Recorded caveats
  • No top-level result JSON. Figures are the gpt_6_astra_max object in the CAD-001 comparison artifact. The Jev-plus-Astra run in that same artifact is not this row. 229 native constituent solids are recorded. That is not a motion-contract pass.
  • Clock is runs.gpt_6_astra_max.wall_clock_seconds. Cost is published_rate_estimate_usd. actual_charged_cost_usd is null. No document URL is published.
  • Source: runs.gpt_6_astra_max.tokens at harness commit 9b7c614dbb5103cc373fdf6275b5c4efc10d8f0e.
Visible evidence gaps
  • campaign evidence; no top-level harness ledger
  • artifact-sourced metrics where a comparison record supplied them
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 Astra
RemakeBenchResearch console