Qwen 3.8 Flash · Apple-stem manipulation presentation recording poster
GLM was the only pass. Gemini lifted the apple using disallowed finger surfaces and therefore did not pass. Compare all four models
Campaign evidence recordPublic result
Qwen 3.8 Flash

Apple-stem manipulation — Qwen 3.8 Flash · Launch 005 campaign evidence

Use the robot-hand fixture to lift the apple by its stem while obeying the allowed-contact-surface rule.

Max reasoningHeadline result
Workflow cost
$0.2397
Wall-clock
1h 28m 21.1s
Processed tokens
Unavailable
Record state
campaign_evidence_not_harness_ledger
Public summary

Qwen 3.8 Flash Apple-stem manipulation campaign-evidence record: 1h 28m 21.1s wall clock and $0.2397 displayed cost.

Run identity and stack
  • Result ID: apple-stem-qwen3-8-flash-launch-005-campaign-evidence
  • Technical model: Qwen 3.8 Flash
  • Provider: Unavailable
  • Client: Unavailable
  • Stack: Launch 005 video campaign evidence
  • Stack: Exact Apple-stem manipulation presentation clip
  • Stack: Not generated from the public Harness ledger
  • Stack: Search alias: Apple Stem
  • Stack: Search alias: Apple-stem manipulation
  • Stack: Search alias: Apple Stem Grasp
Cost basis
  • Exact displayed cost from the owner-supplied Launch 005 media handoff.
  • Campaign clip metric scope; not a substitute for a missing provider invoice or Harness ledger.
Validation evidence
  • Result: Campaign judgment: did not pass
  • Verification: Video-campaign judgment; no machine-readable Harness validator was supplied.
Recorded caveats
  • Campaign-evidence record; this attempt was not generated from the public Harness ledger.
  • Provider, client, token usage, prompt hash, validator output, and attempt number are unavailable unless separately disclosed.
  • The published clip is direct visual evidence, not standardized FPS or hardware evidence.
Visible evidence gaps
  • public Harness ledger provenance
  • provider/client receipt unless separately disclosed
  • token usage
  • prompt hash
  • machine-readable validator output
  • repeat-run evidence
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