BuilderBuilder
GPT-6 Astra · Flooded neo-gothic city
The episode says Sonnet took almost an hour, Astra was quick, and the host had this test on high rather than max. Sonnet's ledger reasoning is max and its elapsed time is 56m 44s. Opus has three partial high attempts. The latest partial attempt is the primary row and the earlier partials stay attached. No attempt is marked passed. Shader capture Unaided model Open this round
Measured harness ledgerPublic result
GPT-6 Astra

Neo-gothic storm city shader — GPT-6 Astra Max

Generate a TWIGL-style neo-gothic storm city shader with legible atmosphere, architecture, lighting, water, and motion.

Max reasoningHeadline result
Workflow cost
$4.86
Wall-clock
17m 06s wall-clock
Processed tokens
2.40M processed
Record state
partial_completed_generation_measurement_with_twigl_smoke_evidence
Public summary

GPT-6 Astra Max partial_completed_generation_measurement_with_twigl_smoke_evidence ledger: 17m 06s wall-clock, 2.40M processed, and $4.86 Verified Standard API-equivalent token estimate; not a subscription invoice..

Run identity and stack
  • Result ID: neo-gothic-storm-city-gpt-6-astra-max
  • Technical model: gpt-6-astra
  • Provider: OpenAI
  • Client: Codex
  • Stack: OpenAI
  • Stack: Codex
  • Stack: Technical model/configuration: gpt-6-astra
  • Stack: TWIGL fragment shader
  • Stack: Harness v1 neo-gothic storm city prompt
Primary artifact integrity
  • Kind: twigl-classic-webgl1-fragment-shader
  • Path: artifacts/neo-gothic-storm-city-gpt-6-astra-max/source/drowned-cathedral.glsl
  • SHA-256: 7696400bc38b00208b2a667a417c581f0e6d1162e8c593ceb2b1f67ffc57c744
Validation evidence
  • Result: PARTIAL_TWIGL_SMOKE_ONLY
Recorded caveats
  • The result is not strict one-shot: the root task also included an environment wrapper and a later metrics request.
  • The row intentionally uses the isolated completed shader-generation turn; the public metrics projection retains the larger root snapshot and explains the difference.
  • The supplied 30 FPS observation is frame-capped local smoke evidence, not the required initial/optimized FPS pair.
  • No RTX Pro 6000 1080p60 final capture or blind evaluation record is archived.
  • Wall-clock is end-to-end workflow latency. Output includes hidden reasoning, visible response content, code, and tool-call JSON.
Visible evidence gaps
  • strict one-user-turn protocol
  • initial and optimized local FPS under identical uncapped conditions
  • RTX Pro 6000 1080p60 final capture with GPU and frame-rate metadata
  • blind evaluation record
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