GLM-5.2: Price, Context, Benchmarks, and Release Details

Z.ai provisional listing open weights · first seen Oct 8, 2026 · score computed Oct 8, 2026, 22:13 UTC

65.7

SI Score (method si-v2-absolute-shrinkage-1)

#30 of 115 ranked

100% confidence — 5 of 7 expected sources in

Coverage 81% of expected source weight · 100% confidence at 80% coverage

Pillar breakdown

Coding 66.2

Math 70.4

Preference 79.4

Reasoning 68.3

Pillar weights: reasoning 30% · math 15% · coding 40% · preference 15%. Benchmark results use fixed 0–100 scales before averaging; incomplete evidence is shrunk toward 50.

Facts

Input price / 1M: not yet reported
Output price / 1M: not yet reported
Context window: 1M
Max output: 131K
Released: Jun 13, 2026
Open weights: Yes
License: not yet reported
Input modalities: text

Benchmark results

38 results

Benchmark	Raw result	Normalized (0–100)	Pillar
arc agi v1 public eval	80.4%	80.4	reasoning
arc agi v1 semi private	77%	77.0	reasoning
arc agi v2 public eval	20.8%	20.8	reasoning
arc agi v2 semi private	22.8%	22.8	reasoning
frontiermath tier 4 v2	29.3%	29.3	math
frontiermath tiers 1 3 v2	54.7%	54.7	math
frontiermath tiers 1 3 v2	59.2%	59.2	math
frontiermath tiers 1 3 v2	42.5%	42.5	math
gpqa diamond	87.9%	87.9	reasoning
gpqa diamond	91.9%	91.9	reasoning
gpqa diamond	71.2%	71.2	reasoning
gpqa diamond	91.2%	91.2	reasoning
hle text only subset	40.5%	40.5	reasoning
hle text only subset tools	54.7%	54.7	reasoning
livebench coding code completion 2026_06_25	80.4%	80.4	coding
livebench coding code generation 2026_06_25	78.9%	78.9	coding
livebench coding javascript 2026_06_25	63.6%	63.6	coding
livebench coding python 2026_06_25	55%	55.0	coding
livebench coding typescript 2026_06_25	36.7%	36.7	coding
livebench math amps hard 2026_06_25	98%	98.0	math
livebench math integrals with game 2026_06_25	76%	76.0	math
livebench math math comp 2026_06_25	96.1%	96.1	math
livebench math olympiad 2026_06_25	89.0%	89.0	math
livebench math simplify 2026_06_25	55.7%	55.7	math
livebench reasoning connections 2026_06_25	94%	94.0	reasoning
livebench reasoning consecutive events 2026_06_25	79.3%	79.3	reasoning
livebench reasoning logic with navigation 2026_06_25	80%	80.0	reasoning
livebench reasoning spatial 2026_06_25	94%	94.0	reasoning
livebench reasoning theory of mind 2026_06_25	75%	75.0	reasoning
livebench reasoning zebra puzzle 2026_06_25	65.5%	65.5	reasoning
lmarena text	1470.4 elo	79.4	preference
otis mock aime 2024 2025	75.6%	75.6	math
otis mock aime 2024 2025	86.4%	86.4	math
otis mock aime 2024 2025	28.9%	28.9	math
swe bench pro	62.1%	62.1	coding
swe bench verified	78.7%	78.7	coding
terminal bench v2 1	82.7%	82.7	coding
terminal bench v2 1	81%	81.0	coding

Normalization uses fixed absolute 0–100 scales for each unit, independently of other models that have a result on each benchmark. Coverage and evidence breadth still affect the composite; compare the evaluation conditions before reading a small score gap as decisive. “lab-reported” marks the provider's own published figure.

Sources: in and pending

in Reported (5)

ARC Prize · arrived Oct 8, 2026
Epoch AI Benchmarking · arrived Oct 8, 2026
Official model cards via models.dev · arrived Oct 8, 2026
LiveBench · arrived Oct 8, 2026
LMArena / Arena · arrived Oct 8, 2026

pending Awaiting (2)

Humanity’s Last Exam · carries 13% of expected weight
Terminal-Bench · carries 6% of expected weight

The confidence % rises as pending sources publish. Some sources never cover some models — that is why 100% confidence arrives at 80% of expected weight, not at full coverage.