Apple Carry — GPT-6 Sol Max

Measured harness ledgerPublic result

Apple Carry — GPT-6 Sol Max

Carry an apple onto a plate on two embodiments: an xArm7 in OpenRoboto's MuJoCo scene, and an FR3 with a Robotiq 2F-85 on SuperDex.

Max reasoningHeadline result

Workflow cost: $0.90
Wall-clock: 10m 21s wall-clock
Processed tokens: 3.13M processed
Record state: judge_pass_independently_replayed

Public summary

GPT-6 Sol Max judge_pass_independently_replayed ledger: 10m 21s wall-clock, 3.13M processed, and $0.90 Standard API-equivalent estimate from the Codex wrapper receipt, not a subscription invoice: this was a ChatGPT-subscription run, $0 was charged and Codex emits no cost field..

Run identity and stack

Result ID: apple-carry-fr3-gripper-gpt-6-sol-max-codex-2026-09-22
Technical model: GPT-6 Sol
Provider: OpenAI Codex
Client: codex-sub wrapper (detached, headless, -l apple-carry-sol -m gpt-6-sol -e max -s danger-full-access), Codex CLI 0.155.1; ran to its own end_turn
Stack: OpenAI Codex
Stack: codex-sub wrapper (detached, headless, -l apple-carry-sol -m gpt-6-sol -e max -s danger-full-access), Codex CLI 0.155.1; ran to its own end_turn
Stack: Technical model/configuration: GPT-6 Sol
Stack: Embodied manipulation fixture
Stack: Harness apple-carry prompt
Stack: Requested tool profile: python-superdex-headless

Cost basis

actual marginal subscription charge usd: $0.00 USD.

Primary artifact integrity

Kind: manifest-verified-artifact
Path: artifacts/apple-carry-fr3-gripper-gpt-6-sol-max-codex-2026-09-22/evidence/model-supplied/seed00/robot.glb
SHA-256: 009696c95e3e3515677b7a65a065c195eb6f34bc72cc453b80c05eb185bbe1a1

Validation evidence

Result: PASS
Path: artifacts/apple-carry-fr3-gripper-gpt-6-sol-max-codex-2026-09-22/validation/judge-replay.public.json
SHA-256: 10ae3beaea04965347c34f6a7255303970d3bc4d0175530249bf5495ca70efdf

Recorded caveats

The judge measures task completion only; there is no visual quality or blind evaluation for this task.
The replay clips are rendered by the operator's viewer from the judge's own trajectory export with a fixed camera; they are presentation evidence, not a measurement.
Wall clock is end-to-end workflow latency including the solver's own simulation and judge runs, not model-only compute.
This was a ChatGPT-subscription Codex run, so there is no dollar receipt: $0 was charged and Codex emits no cost field. API-equivalent is $0.90 at the GPT-6 Sol list rates (corrected 2026-09-24; the original registration wrongly stated no rate was on record).
provenance/orchestrator-results.md is the task project's own result section copied verbatim with absolute local paths replaced.
The raw Codex receipts (events.jsonl, meta.json, prompt and final-message files) carry the session id, the full reasoning transcript and local paths; they are not committed, only their sha256 commitments in provenance/source-receipt-hashes.json.
A Blender Cycles HERO RE-RENDER of this judged replay is pinned in provenance/clips.public.json as presentation: cycles-hero. Poses identical, visual layer only; pipeline ~/hero-render; two RTX 4090 Secure pods (US), 62.8 + 144.6 min, $0.77 + $1.78 + $0.65 aborted batch = $3.21 total. FAIL episodes rendered as-is. It is not a new run, adds no measurement and cannot change a verdict. It is not committed; it is pinned by name, sha256, bytes, frames and duration.
Nudge disclosure: none.
This IS a coding-agent cohort run. It is still not comparable with decision-loop rows.
Five seeds in one session, not twenty. Wilson interval is wide.
Apple-carry's progression is seed 0 only: fewer segments than the five distinct archived seed logs.

Visible evidence gaps

hardware and interpreter receipt from the solver session (only the operator's replay environment is recorded)

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 Builder Return to GPT-6 Sol