GPU TASK Valid result
Legal Stance Retrieval
GPT-6 Luna · Codex · reasoning effort max
Task score (%)9.9
Input Tokens5,445,518including 5,307,722 cached
Output Tokens102,185
Agent time32min
Cost$0.12
Recorded steps68
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
- 2026-10-24T09:36:16.231165Z
- Coding finished
- 2026-10-24T10:09:02.493642Z
- Input (includes cache)
- 5,445,518
- Cached input (subset)
- 5,307,722
- Cache writes (subset)
- 137,601
- Output
- 102,185
- Reasoning (reported separately; not added)
- 69,777
- Recorded responses
- 65
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": 21600.0,
"coding_commit": null,
"effort": "max",
"model": "openai/gpt-6-luna",
"provider": "OpenRouter",
"resources": {
"cpus": 16,
"gpus": 1,
"memory_mb": 32768,
"storage_mb": 102400
},
"run_mode": "official_harbor_full",
"task_commit": "bb34e915be5c2b2948146c2d7862cb58652d6300",
"verifier_commit": "a1ede723abe0add54045301f49a187e11339c134",
"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
legal-stance-retrieval-gpt-6-luna-max-686b15bdc24d- Source trajectory SHA-256
c1d299684d40d8bcc29500fd4c0ab32b841ed921a6e17792c3c0756484535390- Final reward SHA-256
ed53163c862e16c8d37538faea6bae795a1bf754e15e8d51ae6250a886c29cbb