Measured harness ledgerPublic result
GPT-6 Astra + Jev 1.13

Dex Cube Turn — GPT-6 Astra + Jev 1.13 Max

Turn one face of a Rubik's cube a quarter turn by finger contact with two dexterous hands in MuJoCo.

Max reasoningHeadline result
Workflow cost
$2.41
Wall-clock
304.0s wall-clock
Processed tokens
Not recorded
Record state
judge_pass_independently_replayed
Public summary

GPT-6 Astra + Jev 1.13 Max judge_pass_independently_replayed ledger: 304.0s wall-clock, Not recorded, and $2.41 Recorded provider cost.

Run identity and stack
  • Result ID: dex-cube-turn-mujoco-gpt-6-astra-max-jev-coupled-codex-2026-09-21
  • Technical model: GPT-6 Astra + Jev 1.13
  • Provider: OpenAI Codex + TypeSafe via OpenRouter
  • Client: codex-sub -m gpt-6-astra -e max; jev_decide.py after every official self-check
  • Stack: OpenAI Codex + TypeSafe via OpenRouter
  • Stack: codex-sub -m gpt-6-astra -e max; jev_decide.py after every official self-check
  • Stack: Technical model/configuration: GPT-6 Astra + Jev 1.13
  • Stack: MuJoCo dexterous-hand fixture
  • Stack: Harness cube-turn prompt
Primary artifact integrity
  • Kind: manifest-verified-artifact
  • Path: artifacts/dex-cube-turn-mujoco-gpt-6-astra-max-jev-coupled-codex-2026-09-21/evidence/model-supplied/robot.glb
  • SHA-256: fc7116eafaadb1c21c39804797cab20876a65654bf9ce323de5c03b9a4e63fb7
Validation evidence
  • Path: artifacts/dex-cube-turn-mujoco-gpt-6-astra-max-jev-coupled-codex-2026-09-21/validation/judge-replay.public.json
  • SHA-256: e121f3f685eeee373ae71f1cffb3cfd1531df12450b2a04ccebefefbf3e27e4a
Recorded caveats
  • NOT COMPARABLE WITH THE CODING-AGENT COHORTS OR WITH BARE DECISION-LOOP ROWS. This is a jev-astra-hybrid run: GPT-6 Astra and Jev 1.13 were coupled under a protocol that is not the one-shot locked-prompt coding-agent track and not the two-call-per-cycle decision-loop track. cohort_eligible is false. Harry's criterion (same verdict as Astra-alone AND cheaper on API-equivalent) is recorded in the track README; it is not a league-table standing.
  • One run per arm. Cannot separate coupling from run-to-run variance.
  • Jev cents are negligible; extra spend is Astra tokens.
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.

RemakeBenchResearch console