Which is better, GLM-5.2 Full Open Source or Kimi K3?

Across 53 shared benchmarks tracked on Vector Wire, GLM-5.2 Full Open Source leads on 5 and Kimi K3 leads on 48 — Kimi K3 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

GLM-5.2 Full Open Source vs Kimi K3

Z.aivsMoonshot53 shared benchmarks5–48 head-to-head

Benchmark	GLM-5.2 Full Open Source	Kimi K3
AA Agentic Index	45.7	54.3
AA Intelligence	52.6	60
AA-LCR	76.7	82.7
AA-Omniscience	4.4	46
Agents' Last Exam	23.8	28.3
aime_2026	99.2	96.7
apex_agents	35.6	41
apexAgents	33.7	41.3
arc_agi_1	77	65.7
ARC-AGI-2	22.8	12.4
arena_elo	1471	1489
arena_text_factuality	1460	1471
Artificial Analysis Coding Index	68.8	76.2
AutomationBench (Public)	12.9	30.8
coding_arena_elo	1582	1674
critpt	20.9	23.4
DeepSWE	46.2	67.5
DeepSWE 1.1	44	69
DSBench-FullStack	61.8	73.7
DSBench-Hard	54.5	63
gdpval	50.3	59.4
GDPval-AA v2	1514	1687
GDPval-AA v2 (Elo)	1510	1686
GPQA Diamond	91.2	93.5
HLE	54.7	56
HLE (wo / w tools)	54.7	56
hmmt_feb_2026	92.5	97
itbenchSre	42.7	47.7
JobBench	43.4	54.3
livebench_agentic_coding	51.8	62.2
livebench_coding	79.7	81.5
livebench_data_analysis	73.7	78.7
livebench_instruction_following	62.3	71.4
livebench_language	76.2	85.5
livebench_math	89.8	84.4
livebench_reasoning	78.6	90.7
MCP Atlas	82.6	84.2
MLS Bench Lite	40.4	48.3
officeqa_pro	41.4	63.3
OmniScience Accuracy	24.3	47.6
OmniScience Non-Hallucination	73.7	49.1
PostTrainBench	34.3	36.6
Program Bench	63.7	77.8
ResearchRubrics	71.1	76.2
scicode	50.5	58.7
simplebench	58.8	60.7
simpleqa_verified	38.1	42.7
SpreadsheetBench 2	28.1	34.8
SWE-Marathon	13	42
Tau 3 Banking	26.8	33.4
TauBench V3 - Banking	34.6	46
Terminal-Bench 2.1	82.7	88.3
toolathlon	48.2	73.2

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.