Which is better, GLM-5.1 or GPT-5.4?

Across 54 shared benchmarks tracked on Vector Wire, GLM-5.1 leads on 13 and GPT-5.4 leads on 41 — GPT-5.4 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

GLM-5.1 vs GPT-5.4

Z.aivsOpenAI54 shared benchmarks13–41 head-to-head

Benchmark	GLM-5.1	GPT-5.4
AA Agentic Index	66	58.2
AA Intelligence	41	53.1
AA-LCR	68	77.7
AA-Omniscience	0.8	5.8
AgentWorldBench - Android	59.1	60
AgentWorldBench - MCP	67.6	70.1
AgentWorldBench - OS	59.1	68.6
AgentWorldBench - Overall	51.3	58.3
AgentWorldBench - Search	22.5	37.3
AgentWorldBench - SWE	52.1	66.3
AgentWorldBench - Terminal	47.3	53.7
AgentWorldBench - Web	51.5	51.8
aime_2026	95.8	99.2
Apex (Pass@1)	11.5	54.1
Apex Shortlist (Pass@1)	72.4	78.1
arena_elo	1468	1477
arena_text_factuality	1458	1476
Artificial Analysis Coding Index	55.8	71.1
browsecomp	79.3	82.7
browsecomp_with_context_manager	79.3	82.7
coding_arena_elo	1510	1457
critpt	4.6	23.4
cybergym	68.7	66.3
frontiermath_tier_4	12.5	27.1
gdpval	49.5	50.1
GDPval-AA (Elo)	1535	1674
GPQA Diamond	86.8	93
HLE	52.3	43.7
HLE (with tools)	52.3	52.1
hmmt_feb_2026	89.4	97.7
hmmt_nov_2025	94	95.8
IFBench	76.3	73.9
imo_answer_bench	83.8	91.4
itbenchSre	40.3	34.5
livebench	70.2	80.3
MCP Atlas	75.6	70.6
MCPAtlas Public (Pass@1)	71.8	67.2
MMLU-Pro	86	87.5
nl2repo	42.7	41.3
OmniScience Accuracy	25.2	50.9
OmniScience Non-Hallucination	70.1	17.4
scicode	43.8	56.6
simpleqa_verified	38.1	45.3
SWE-bench Multilingual	73.3	71.7
SWE-bench Pro	58.4	59.1
SWE-QA	72.7	81.3
TauBench V3 - Banking	13.6	39.6
Terminal-Bench 2.0	69	75.1
Terminal-Bench 2.1	63.5	78.3
Terminal-Bench Hard	43.2	57.6
tool_decathlon	40.7	54.6
toolathlon	40.7	54.6
τ²-Bench Telecom (AA run)	97.7	87.1
τ³-Bench	70.6	72.9

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.