Model result, rank 13 of 27

GLM‑5.2

OpenRouter · z-ai/glm-5.2 · maximum 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 #17 of 27

77.35

SWE MVPlane rank #1 of 27

89.58

Hard Intelligencelane rank #18 of 27

74.58

GLM‑5.2
Cohort median
Cohort best

Overall score80.50

Rank #13

Full / Agentic77.35

Full rank #17

SWE MVP89.58

SWE rank #1

Hard Intelligence74.58

Hard rank #18

Measured cost$2.768

100.0% reliability

Overall

All-around publication view

Score80.50

Formulamean(Full, SWE, Hard Intelligence)

BasisOpenRouter · z-ai/glm-5.2 · maximum reasoning · Full + SWE + Hard measured

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

Lane 01

Full / Agentic benchmark

Final77.35

Capability99.37

Agentic93.18

Pass rate100.0%

Prompts43

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

Lane 02

Software engineering MVP

SWE score89.58

Focused final71.92

Capability89.58

Daily driver65.89

Prompts24

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

Lane 03

Hard Intelligence diagnostic

Hard score74.58

Active inquiry17.50

Online adaptation98.13

Self-repair82.71

Authority integrity100.00

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

Telemetry

Runtime economics

Total cost$2.768

Cost / scored item$0.037

Seconds / timed item91.19s

Runtime coverage100.0%

Recorded tokens / item22.4k

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 SWE lane than in the Full/Agentic lane; the overall score is therefore shown with both component lanes visible. Hard Intelligence score is 74.58 and contributes to the overall score alongside Full/Agentic and SWE. GLM‑5.2 used OpenRouter maximum reasoning across the public benchmark suites. SWE score reflects all prompt outcomes from the public software-engineering suite.