Search-SWE
GPU TASK Valid result

Composed Image Search

GPT-6 Luna · Codex (v0.164.0) · reasoning effort max

Task score (%)36.3
Input Tokens6,697,157including 6,469,145 cached
Output Tokens142,581
Agent time49min
Cost$0.16
Recorded steps64

Cost is estimated from recorded agent tokens at official API rates in USD (2026-11-01; DeepSeek off-peak pricing).

Loading trajectory…

Select this tab to browse the saved files.

Coding time & token usage
Original agent completion
normal
Coding started
2026-10-28T17:00:08.097052Z
Coding finished
2026-10-28T17:49:44.854648Z
Input (includes cache)
6,697,157
Cached input (subset)
6,469,145
Cache writes (subset)
182,693
Output
142,581
Reasoning (reported separately; not added)
106,553
Recorded responses
61

Input includes cached input; output is counted once. Auxiliary reasoning counts are shown as recorded. Regrading does not repeat coding usage.

Select this tab to browse the saved files.

Configuration

Evaluation settings summary
{
  "agent": "Codex",
  "agent_timeout_sec": null,
  "coding_commit": null,
  "effort": "max",
  "model": "openai/gpt-6-luna",
  "provider": "OpenRouter",
  "resources": null,
  "run_mode": "official_harbor_full",
  "task_commit": "1d73c6f9c4a90ba922b7572d9395a6f40661bbb2",
  "verifier_commit": "1d73c6f9c4a90ba922b7572d9395a6f40661bbb2",
  "version": "0.164.0"
}

Select this tab to browse the saved files.

Resolved configuration

Select this tab to browse the saved files.

Trial log

Select this tab to browse the saved files.

Exceptions

Select this tab to browse the saved files.

Select this tab to browse the saved files.

Trajectory downloads & provenance

Public text redacts private values and normalizes host paths. Source hashes identify the original records.

Trial ID
composed-image-search-gpt-6-luna-max-4800b458acbd
Source trajectory SHA-256
64eee443332d789f200ff35cbec859332c99bed5348ac40c8fe7c8985d215221
Final reward SHA-256
e3f13f68b70919454f2866072de8ead8884a2cbd950f2f7f6abe12a73f4996fc