Which is better, DeepSeek-V3.2 or GLM-5.2 Full Open Source?

Across 35 shared benchmarks tracked on Vector Wire, DeepSeek-V3.2 leads on 2 and GLM-5.2 Full Open Source leads on 33 — GLM-5.2 Full Open Source leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

DeepSeek-V3.2 vs GLM-5.2 Full Open Source

DeepSeekvsZ.ai35 shared benchmarks2–33 head-to-head

Benchmark	DeepSeek-V3.2	GLM-5.2 Full Open Source
AA Agentic Index	39.8	45.7
AA Intelligence	32.8	52.6
AA-LCR	70.7	76.7
AA-Omniscience	-22.5	4.4
aime_2026	95.1	99.2
apexAgents	14.5	33.7
arc_agi_1	57	77
ARC-AGI-2	4	22.8
arena_text_factuality	1433	1460
Artificial Analysis Coding Index	44.2	68.8
coding_arena_elo	1368	1582
critpt	2.9	20.9
gdpval	18.8	50.3
GPQA Diamond	84	91.2
HLE	40.8	54.7
HLE (with tools)	40.8	54.7
HMMT 2025	90.2	94.4
hmmt_feb_2026	84.1	92.5
hmmt_nov_2025	90.2	94.4
IFBench	61	73.3
imo_answer_bench	78.3	91
MCP Atlas	62.2	82.6
OmniScience Accuracy	33	24.3
OmniScience Non-Hallucination	17.3	73.7
ResearchRubrics	55.8	71.1
scicode	39	50.5
simpleqa_verified	27.5	38.1
SWE-bench Pro	15.6	62.1
TauBench V3 - Banking	18.8	34.6
Terminal-Bench 2.1	46.8	82.7
Terminal-Bench Hard	35.6	50.8
tool_decathlon	35.2	48.2
toolathlon	35.2	48.2
τ²-Bench Telecom (AA run)	90.6	99.1
τ³-Bench	69.2	26.8

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.