THE FORGE RANKINGS·2026-08-30

GLM-5.2

Z.ai (Zhipu AI)
14
RANK of 37 ranked
FERROX INDEX71.1 weighted mean of category positions
GRADEB above the field median
EVIDENCE9 / 32 benchmarks measured
QUALITYFULL every category rests on independent benchmarks

Not separable from Claude Opus 4.6 and Gemini 3.1 Pro Preview at the measured 1.78 point band. The order is printed; the gap is not claimed.

No measurement in Preference. That weight was redistributed, which is the same as assuming this model would have scored its own average there. It is an assumption, not a measurement.

Compared to what

Each rail is one classification. Every ranked model measured on it is a tick; this model is the orange marker. A hatched rail means this model has no measurement in that classification and its weight was redistributed.

Agents

24of 37 measured
36.0
13.1median 38.963.0

Coding

33of 37 measured
37.0
22.4median 50.070.7

Reasoning

20of 37 measured
62.1
18.1median 62.893.0

Preference

no qualifying measurement
53.1median 71.882.5

Every measurement, and the field behind it

One card per benchmark this model has been measured on. The rail shows where it sits against every other ranked model measured on the same benchmark. Where the evaluator keys on a harness, the harness is named, because the same model scores differently under a different one.

ARC-AGI-2

18/29
22.8% best 88.3% · Claude Opus 5
ARC Prize CC-BY-4.0

APEX Agents

8/28
35.6% best 45.0% · Claude Fable 5
APEX CC-BY-4.0

LMArena Agent

10/21
0.059 best 0.127 · Claude Opus 5
LMArena CC-BY-4.0

SWE-bench Verified (Epoch's own run)

2/21
78.7% best 83.5% · Claude Opus 4.7
Epoch AI CC-BY-4.0 max

FrontierCode

15/17
24.5% best 53.5% · Claude Fable 5
cognition.com CC-BY-4.0 mini-swe-agent none

Not measured

23/32
  • FrontierMath
  • LMArena Text (style-controlled)
  • Humanity's Last Exam
  • Terminal-Bench 2
  • LMArena WebDev
  • METR time horizons
  • Terminal-Bench 4.0
  • Aider Polyglot
  • OSWorld
  • Cybench
  • BrowseComp
  • IMOAnswerBench
  • SWE-bench Multilingual
  • AIME
  • CyberGym
  • HMMT
  • MCP-Atlas
  • SWE-bench Pro
  • Tau2-Bench Airline
  • Tau2-Bench Banking
  • Tau2-Bench Retail
  • Tau2-Bench Telecom
  • Tool-Decathlon
A missing benchmark is not a zero and is never scored as one.

Check it yourself

Every number on this page, with who measured it, the interval they published, the harness it was run under, and where to go and read it. Nothing here was measured by Ferrox Labs.

BenchmarkClassPublishedInterval (normalised) Measured byLicenceHarness / effort
APEX Agents agents 35.6% 28.5 to 42.7 APEX CC-BY-4.0 source
LMArena Agent agents 0.059 33.8 to 39.1 LMArena CC-BY-4.0 source
ALE-Bench coding 1010 not published ALE-Bench (Sakana AI with AtCoder) CC-BY-4.0 / max source
DeepSWE coding 36.3% 31.5 to 41.0 deepswe.datacurve.ai CC-BY-4.0 mini-swe-agent / high source
FrontierCode coding 24.5% not published cognition.com CC-BY-4.0 mini-swe-agent / none source
SWE-bench Verified (Epoch's own run) coding 78.7% 75.0 to 82.4 Epoch AI CC-BY-4.0 / max source
ARC-AGI-2 reasoning 22.8% not published ARC Prize CC-BY-4.0 source
GPQA Diamond (Epoch's own run) reasoning 87.9% 83.3 to 92.4 Epoch AI CC-BY-4.0 / low source
OTIS Mock AIME 2024-2025 reasoning 75.6% 62.9 to 88.3 Epoch AI CC-BY-4.0 / low source

Configurations rolled up: 8. Rule: median observed configuration per benchmark (lower median, always a real measurement). Harnesses seen: mini-swe-agent.

Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e. Grades are positional: position in the measured field, as a percentile of rank among ranked entries, n=37.