Model result, rank 20 of 27

GPT‑6 Luna

ChatGPT Codex subscription · extra-high reasoning · Full + SWE + Hard measured. Public result card with the model’s overall score, lane measurements, runtime and cost telemetry, and the ranking formula.

Compare all models Download public JSON

Figure 1 Lane profile against the cohort Each bar is one measured lane on the shared 0 to 100 scale. The thin mark is the cohort median; the bright mark is the best score any ranked model reached on that lane.

Full / Agenticlane rank #13 of 27

78.57

SWE MVPlane rank #20 of 27

62.59

Hard Intelligencelane rank #13 of 27

83.82

GPT‑6 Luna
Cohort median
Cohort best

Overall score74.99

Rank #20

Full / Agentic78.57

Full rank #13

SWE MVP62.59

SWE rank #20

Hard Intelligence83.82

Hard rank #13

Measured cost$0.070

100.0% reliability

Overall

All-around publication view

Score74.99

Formulamean(Full, SWE, Hard Intelligence)

BasisChatGPT Codex subscription · extra-high reasoning · Full + SWE + Hard measured

The overall score averages the measured major lanes while keeping each source measurement visible.

Lane 01

Full / Agentic benchmark

Final78.57

Capability88.41

Agentic90.36

Pass rate91.4%

Prompts43

This lane captures instruction following, structured behavior, tool discipline, and general agentic reliability.

Lane 02

Software engineering MVP

SWE score62.59

Focused final61.57

Capability53.23

Daily driver64.47

Prompts24

This lane is closer to implementation usefulness: source handling, architecture cleanliness, and deliverable quality.

Lane 03

Hard Intelligence diagnostic

Hard score83.82

Active inquiry99.75

Online adaptation80.00

Self-repair85.54

Authority integrity70.00

Hard Intelligence measures active inquiry, online adaptation, evidence-driven self-repair, and authority/salience integrity.

Telemetry

Runtime economics

Total cost$0.070

Cost / scored item$0.0008

Seconds / timed item13.43s

Runtime coverage100.0%

Recorded tokens / item7.0k

Token coverage100.0%

Cost, time, and token basis are normalized telemetry. They explain tradeoffs; they do not overwrite the capability score yet.

Interpretation

Why the result lands here.

The model is stronger in the Full/Agentic lane than in the SWE lane; the overall score is therefore shown with both component lanes visible. Hard Intelligence score is 83.82 and contributes to the overall score alongside Full/Agentic and SWE. Runs on the ChatGPT Codex subscription; the transport is the Codex Responses route rather than a metered API key. Reasoning effort is extra-high on every lane, matching the published extra-high baselines. Costed at the model’s published Standard API rates ($0.10/1M input, $0.50/1M output) as a comparable basis, not an invoice: the subscription bills no marginal per-token charge. Hard Intelligence is a public diagnostic, not a hidden official score. Token median and P90 per scored item are intentionally omitted: their derivation could not be reproduced on the same basis as the published rows.