Model result, rank 9 of 27

GLM 5.3 Flash

z.ai Coding Plan · glm-5.3-flash · maximum reasoning with thinking enabled · 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 #19 of 27

76.49

SWE MVPlane rank #16 of 27

71.68

Hard Intelligencelane rank #3 of 27

95.78

GLM 5.3 Flash
Cohort median
Cohort best

Overall score81.32

Rank #9

Full / Agentic76.49

Full rank #19

SWE MVP71.68

SWE rank #16

Hard Intelligence95.78

Hard rank #3

Measured cost$0.117

99.2% reliability

Overall

All-around publication view

Score81.32

Formulamean(Full, SWE, Hard Intelligence)

Basisz.ai Coding Plan · glm-5.3-flash · maximum reasoning with thinking enabled · Full + SWE + Hard measured

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

Lane 01

Full / Agentic benchmark

Final76.49

Capability95.02

Agentic87.51

Pass rate94.3%

Prompts43

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

Lane 02

Software engineering MVP

SWE score71.68

Focused final64.26

Capability66.41

Daily driver62.77

Prompts24

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

Lane 03

Hard Intelligence diagnostic

Hard score95.78

Active inquiry99.75

Online adaptation97.97

Self-repair85.39

Authority integrity100.00

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

Telemetry

Runtime economics

Total cost$0.117

Cost / scored item$0.0014

Seconds / timed item35.45s

Runtime coverage100.0%

Recorded tokens / item8.6k

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 result is best read as a balanced benchmark entry: one overall score plus the lane measurements that produced it. Hard Intelligence score is 95.78 and contributes to the overall score alongside Full/Agentic and SWE. Runs on the z.ai Coding Plan subscription rather than a metered API key. Reasoning effort is max with thinking enabled on every lane, which is the strongest setting this route exposes. Costed at z.ai’s published rates ($0.15/1M input, $0.50/1M output, $0.03/1M cached input) as a comparable basis, not an invoice. Implicit prompt caching is billed at the cached rate. Thinking cannot be disabled on this route, so reasoning tokens dominate the output; on the Hard diagnostic they were 93.9% of all emitted output tokens. 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.