Measured harness ledgerPublic result
Grok 4.7Apple Stem — Grok 4.7 xhigh
Pinch an apple by the stem and lift it with a Wuji hand on an FR3 arm in SuperDex, without contacting the apple body.
xhigh reasoningHeadline result
- Workflow cost
- $5.18
- Wall-clock
- 118m 14s wall-clock
- Processed tokens
- 16.34M processed
- Record state
- judge_fail_independently_replayed
Public summary
Grok 4.7 xhigh judge_fail_independently_replayed ledger: 118m 14s wall-clock, 16.34M processed, and $5.18 Provider-recorded usage estimate, not an itemized subscription cash charge.
Run identity and stack
- Result ID: dex-apple-stem-superdex-grok-4.7-xhigh-2026-09-22
- Technical model: Grok 4.7
- Provider: xAI Grok
- Client: grok-sub wrapper (detached, headless, sandbox off), grok-4.7 at xhigh; ran to its own end_turn
- Stack: xAI Grok
- Stack: grok-sub wrapper (detached, headless, sandbox off), grok-4.7 at xhigh; ran to its own end_turn
- Stack: Technical model/configuration: Grok 4.7
- Stack: SuperDex fingertip fixture
- Stack: Harness apple-stem prompt
- Stack: Requested tool profile: python-superdex-headless
Cost basis
- usage cost reported by xAI for the session (grok-sub receipt line, confirmed by the wrapper run's meta.json)
Primary artifact integrity
- Kind: open-loop-200hz-actuator-log
- Path: artifacts/dex-apple-stem-superdex-grok-4.7-xhigh-2026-09-22/evidence/model-supplied/ctrl.npy
- SHA-256: f3c379746fd410194be210b2ea6b6eec3e6f3f075ce542781e478d9892458f6f
Validation evidence
- Result: PASS
- Path: artifacts/dex-apple-stem-superdex-grok-4.7-xhigh-2026-09-22/validation/judge-replay.public.json
- SHA-256: 269dbd28746a5dbd39e5d71786c0c67033ba512fd007f2486761723d6defdd42
- Validator SHA-256: c735d44ca72534f8058a7e1061ccf775f6e9f180cef26200d97a402f3de7403e
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.
- Cost is the xAI-reported session total from the grok-sub receipt line.
- provenance/orchestrator-results.md is the task project's own result section copied verbatim with absolute local paths replaced.
- The raw grok-sub receipts carry the session id, the 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 EU-RO-1 + EUR-IS-2, 52 + 154 min, $0.64 + $1.90 + $0.25 aborted batch = $2.79 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.
- The 2026-09-21 launch was operator-killed at budget for slow internet (exit 143) and is archived at runs/grok-4.7-xhigh-interrupted-2026-09-21/; it is pinned by hash in provenance/interrupted-2026-09-21.json and is not scored.
- The counted 2026-09-22 re-run's final self-check was killed under a host load average near 600 (six parallel solvers + sims + leaked headless Chrome). That is an environment caveat on this result, not a model-competence claim beyond the submitted 3 s hold.
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.