Search-SWE
CPU TASK Valid result

Code Context Retrieval

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

Task score (%)16.7
Input Tokens51,839,738including 51,348,852 cached
Output Tokens353,680
Agent time1h 35min
Cost$1.23
Recorded steps175

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
2026-10-24T09:36:29.472511Z
Coding finished
2026-10-24T11:11:48.067305Z
Input (includes cache)
51,839,738
Cached input (subset)
51,348,852
Cache writes (subset)
463,014
Output
353,680
Reasoning (reported separately; not added)
268,145
Recorded responses
172

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": "saved_outputs_production_rescore",
  "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
code-context-retrieval-gpt-6-luna-max-11c1e408bc2a
Source trajectory SHA-256
e84d4249dd729e00b880b781196368ca7619d219995c52551e739681f78245b5
Final reward SHA-256
c79b47cb3fca53e7c97d8a827e4f36a7cd7af58fadd9868b8523a4c31fe840a9