Head-to-head comparison

GPT-5.4 vs. GLM-5.2

Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.

shared result series: 165
distinct benchmarks: 16
latest retrieval: 2026-09-04

First modelSecond model

Specifications

General specifications

Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.

Specifications for GPT-5.4 and GLM-5.2
Attribute	GPT-5.4	GLM-5.2
Provider	OpenAI	Z.ai
Released	Mar 2026	Jun 16, 2026
Availability	Active	Open Source
Model type	Proprietary	Open Weights
Parameters	Not published	744B, 40B active
Architecture	Not published	Mixture of Experts
Context window	1,050,000 tokens	1,000,000 tokens
Knowledge cutoff	August 31, 2025	Not published
Technical sources	Model	Model, Context

Pricing

API pricing compared

The table compares official standard API rates per 1M tokens, including published cache pricing. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.

API pricing for GPT-5.4 and GLM-5.2
Price type	GPT-5.4	GLM-5.2
API input	$2.5 ≤272K / $5 >272K	$1.4
API output	$15 ≤272K / $22.5 >272K	$4.4
Cache read	$0.25 ≤272K / $0.5 >272K	$0.26
Cache write (5 min.)	Not listed	Not listed
Cache write (1 hr.)	Not listed	Not listed
Price verified	07/30/2026 Source	07/30/2026 Source

GPT-5.4GLM-5.2

API inputUSD per 1M tokens, base tier

GPT-5.4$2.5

GLM-5.2$1.4

API outputUSD per 1M tokens, base tier

GPT-5.4$15

GLM-5.2$4.4

Cache readUSD per 1M tokens, base tier

GPT-5.4$0.25

GLM-5.2$0.26

Key to the marks:better valuebetter, but not conclusivetie

Pricing by provider

Direct developer pricing and provider endpoint tariffs routed through OpenRouter are listed separately. Prices are in USD per 1M tokens. Context-dependent direct tiers plus regional, Flex, Priority, and other OpenRouter tariffs remain individually identifiable.

Provider pricing for GPT-5.4 and GLM-5.2
Model	Provider	Purchase route and tariff	Input per 1M tokens	Output per 1M tokens	Source
GPT-5.4	OpenAI	Direct from developerUp to 272,000 context tokens	$2.5	$15	Source07/30/2026
	OpenAI	Direct from developerAbove 272,000 context tokens	$5	$22.5	Source07/30/2026
	Amazon Bedrock	Via OpenRouteramazon-bedrock/us-east-1	$2.75	$16.5	Source09/06/2026
	Azure	Via OpenRouterazure	$2.5	$15	Source09/06/2026
	Azure	Via OpenRouterazure/eu	$2.75	$16.5	Source09/06/2026
	Azure	Via OpenRouterazure/us	$2.75	$16.5	Source09/06/2026
	OpenAI	Via OpenRouteropenai	$2.5	$15	Source09/06/2026
	OpenAI	Via OpenRouteropenai/fast	$5	$30	Source09/06/2026
	OpenAI	Via OpenRouteropenai/flex	Lowest listed price: $1.25	Lowest listed price: $7.5	Source09/06/2026
GLM-5.2	Z.ai	Direct from developerStandard API	$1.4	$4.4	Source07/30/2026
	Alibaba	Via OpenRouteralibaba/fast	$2.31	$7.26	Source09/06/2026
	Alibaba	Via OpenRouteralibaba/fp8	$0.966	$3.036	Source09/06/2026
	Ambient	Via OpenRouterambient/fp8	$0.6	$2	Source09/06/2026
	AtlasCloud	Via OpenRouteratlas-cloud/fp8	$0.938	$2.948	Source09/06/2026
	Baidu	Via OpenRouterbaidu/fp8	$0.357	$1.122	Source09/06/2026
	BaseTen	Via OpenRouterbaseten/fast#1	$2.1	$6.6	Source09/06/2026
	BaseTen	Via OpenRouterbaseten/fast#2	$2.1	$6.6	Source09/06/2026
	BaseTen	Via OpenRouterbaseten/fp8#1	$1.4	$4.4	Source09/06/2026
	BaseTen	Via OpenRouterbaseten/fp8#2	$1.4	$4.4	Source09/06/2026
	Cloudflare	Via OpenRoutercloudflare	$1.4	$4.4	Source09/06/2026
	CoreWeave	Via OpenRoutercoreweave/fp4	$0.76	$2.42	Source09/06/2026
	Crusoe	Via OpenRoutercrusoe/fp8	$1.4	$4.4	Source09/06/2026
	Decart	Via OpenRouterdecart/fp4	$0.684	$2.28	Source09/06/2026
	DeepInfra	Via OpenRouterdeepinfra/fp4	$0.4875	$1.56	Source09/06/2026
	DigitalOcean	Via OpenRouterdigitalocean	$0.7	$2.2	Source09/06/2026
	Fireworks	Via OpenRouterfireworks	$1.4	$4.4	Source09/06/2026
	Fireworks	Via OpenRouterfireworks/fast	$2.1	$6.6	Source09/06/2026
	Fireworks	Via OpenRouterfireworks/fast-us	$2.1	$6.6	Source09/06/2026
	Friendli	Via OpenRouterfriendli	$1.4	$4.4	Source09/06/2026
	GMICloud	Via OpenRoutergmicloud/fp8	$1.05	$3.3	Source09/06/2026
	Inceptron	Via OpenRouterinceptron/fp4	$1.25	$2.99	Source09/06/2026
	Mistral	Via OpenRoutermistral	$1.4	$4.4	Source09/06/2026
	Mistral	Via OpenRoutermistral/eu	$1.54	$4.84	Source09/06/2026
	Mistral	Via OpenRoutermistral/zdr	$1.4	$4.4	Source09/06/2026
	Novita	Via OpenRouternovita/fp8	$0.5586	$1.7556	Source09/06/2026
	Parasail	Via OpenRouterparasail/fp4	$1.4	$4.4	Source09/06/2026
	Phala	Via OpenRouterphala/fp8	$1.26	$3	Source09/06/2026
	SiliconFlow	Via OpenRoutersiliconflow/fp8	$1.19	$3.74	Source09/06/2026
	StreamLake	Via OpenRouterstreamlake/fp8	Lowest listed price: $0.343	Lowest listed price: $1.078	Source09/06/2026
	Together	Via OpenRoutertogether	$1.4	$4.4	Source09/06/2026
	Venice	Via OpenRoutervenice/fp8	$1.4	$4.4	Source09/06/2026
	Z.AI	Via OpenRouterz-ai/fp8	$1.4	$4.4	Source09/06/2026

Performance

Shared benchmarks

Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.

Capability profile from matched benchmarks

Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.

GPT-5.4

GLM-5.2

Coding: 77.4 / 95.5 (5 benchmark families)

Reasoning: 100 / 93.4 (1 benchmark family)

Knowledge: 100 / 99.1 (1 benchmark family)

Finance: 99.4 / 99.6 (2 benchmark families)

Legal: 50 / 98.9 (2 benchmark families)

Agentic tasks: 100 / 87.2 (1 benchmark family)

Cybersecurity: 93.9 / 99.9 (1 benchmark family)

Health: 96.4 / 99.4 (2 benchmark families)

17 matched rows from 15 benchmark families were considered. The radar shows 8 of 8 comparable categories.

GPT-5.4GLM-5.2

Code MigrationOverall. Higher is better. No documented setup difference

GPT-5.434.984 %

GLM-5.237.87 %

CorpFin v2Overall. Higher is better. No documented setup difference

GPT-5.465.268 %

GLM-5.266.123 %

CyberBenchOverall. Higher is better. No documented setup difference

GPT-5.472.881 %

GLM-5.277.359 %

GPQA DiamondOverall. Higher is better. No documented setup difference

GPT-5.491.666 %

GLM-5.285.606 %

Harvey's Legal Agent BenchmarkOverall · Task fully resolved. Higher is better. No documented setup difference

GPT-5.40 %

GLM-5.27.083 %

LegalBenchOverall. Higher is better. No documented setup difference

GPT-5.486.044 %

GLM-5.284.073 %

LiveCodeBench v6Overall. Higher is better. No documented setup difference

GPT-5.484.141 %

GLM-5.269.5 %

LMArena Text, Style ControlOverall. Higher is better. Test setup not documented as identical

GPT-5.41,465.527

GLM-5.21,471.886

MedCodeOverall. Higher is better. No documented setup difference

GPT-5.441.292 %

GLM-5.240.771 %

MedScribeOverall. Higher is better. No documented setup difference

GPT-5.477.549 %

GLM-5.283.534 %

MMLU-ProOverall. Higher is better. No documented setup difference

GPT-5.487.482 %

GLM-5.286.714 %

ProgramBenchOverall. Higher is better. No documented setup difference

GPT-5.40 %

GLM-5.20.5 %

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for GPT-5.4 and GLM-5.2
Benchmark and task	GPT-5.4	GLM-5.2	Source
Code MigrationOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	34.984 %	Winner37.87 %	Vals AI Source details Data as of: 09/03/2026
CorpFin v2Overall Test details No documented difference in the test setupVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	65.268 %	Winner66.123 %	Vals AI Source details Data as of: 08/12/2026
CyberBenchOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	72.881 %	Winner77.359 %	Vals AI Source details Data as of: 09/01/2026
GPQA DiamondOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner91.666 %	85.606 %	Vals AI Source details Data as of: 09/01/2026
Harvey's Legal Agent BenchmarkOverall · Task fully resolved Test details No documented difference in the test setupVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	0 %	Winner7.083 %	Vals AI Source details Data as of: 09/03/2026
LegalBenchOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner86.044 %	84.073 %	Vals AI Source details Data as of: 09/01/2026
LiveCodeBench v6Overall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner84.141 %	69.5 %	Vals AI Source details Data as of: 09/01/2026
LMArena Text, Style ControlOverall Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 63,615 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 33,758 anonymous pairwise votes.	1,465.527 95% confidence interval: ±3.807The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	Higher measured value1,471.886 95% confidence interval: ±4.729The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MedCodeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner41.292 %	40.771 %	Vals AI Source details Data as of: 09/03/2026
MedScribeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	77.549 %	Winner83.534 %	Vals AI Source details Data as of: 09/03/2026
MMLU-ProOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner87.482 %	86.714 %	Vals AI Source details Data as of: 09/01/2026
ProgramBenchOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	0 %	Winner0.5 %	Vals AI Source details Data as of: 09/01/2026
SkillsBenchOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Harness: OpenHands GLM-5.2: Harness: OpenHands	Winner51.711 %	45.077 %	Vals AI Source details Data as of: 09/01/2026
SWE-bench VerifiedOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	78.2 %	Winner82.8 %	Vals AI Source details Data as of: 09/01/2026
TaxEval v2Overall Test details No documented difference in the test setupVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner73.958 %	73.344 %	Vals AI Source details Data as of: 09/01/2026
Vibe Code Bench v1.1Overall Test details No documented difference in the test setupVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Harness: OpenHands GLM-5.2: Harness: OpenHands	Winner67.421 %	63.961 %	Vals AI Source details Data as of: 09/03/2026
Code MigrationCLI Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	23.788 %	Higher measured value27.16 %	Vals AI Source details Data as of: 08/09/2026
CorpFin v2Exact Pages Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value64.103 %	63.636 %	Vals AI Source details Data as of: 08/06/2026
CyberBenchPatch Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	Winner81.356 %	81.034 %	Vals AI Source details Data as of: 08/03/2026
GPQA DiamondFew-Shot CoT Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value91.414 %	83.333 %	Vals AI Source details Data as of: 08/09/2026
Harvey's Legal Agent BenchmarkAntitrust Competition · Criteria passed Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: criteria pass rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	76.449 %	Higher measured value87.319 %	Vals AI Source details Data as of: 08/07/2026
LegalBenchConclusion Tasks Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value88.939 %	88.587 %	Vals AI Source details Data as of: 08/09/2026
LiveCodeBench v6Easy Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value99.068 %	97.205 %	Vals AI Source details Data as of: 08/09/2026
LMArena Text, Style ControlOverall Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-high, Reasoning: high, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 60,657 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 33,758 anonymous pairwise votes.	Higher measured value1,476.488 95% confidence interval: ±3.881The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	1,471.886 95% confidence interval: ±4.729The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProBiology Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	92.887 %	Higher measured value94.003 %	Vals AI Source details Data as of: 08/09/2026
ProgramBenchAlmost Resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 64,000 tokens	7 %	Higher measured value9.5 %	Vals AI Source details Data as of: 08/09/2026
SkillsBenchNo Skills Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Harness: OpenHands, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Harness: OpenHands, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	Higher measured value48.905 %	32.223 %	Vals AI Source details Data as of: 08/07/2026
SWE-bench Verified<15 min fix Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	No direct winner88.144 %	No direct winner88.144 %	Vals AI Source details Data as of: 08/08/2026
TaxEval v2Correctness Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value64.922 %	62.061 %	Vals AI Source details Data as of: 08/09/2026
Code MigrationCOBOL Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	68.571 %	Higher measured value70 %	Vals AI Source details Data as of: 08/09/2026
CorpFin v2Max Fitting Context Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	64.918 %	Higher measured value68.415 %	Vals AI Source details Data as of: 08/06/2026
CyberBenchPoC Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05	64.407 %	Winner73.684 %	Vals AI Source details Data as of: 08/03/2026
GPQA DiamondZero-Shot CoT Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value91.919 %	87.879 %	Vals AI Source details Data as of: 08/09/2026
Harvey's Legal Agent BenchmarkAntitrust Competition · Task fully resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 08/07/2026
LegalBenchInterpretation Tasks Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value83.921 %	82.17 %	Vals AI Source details Data as of: 08/09/2026
LiveCodeBench v6Hard Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value67.714 %	37.143 %	Vals AI Source details Data as of: 08/09/2026
LMArena Text, Style ControlBusiness, management and finance Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 13,060 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 6,609 anonymous pairwise votes.	Higher measured value1,474.871 95% confidence interval: ±6.534The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	1,460.145 95% confidence interval: ±8.046The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProBusiness Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	91.762 %	Higher measured value91.888 %	Vals AI Source details Data as of: 08/09/2026
ProgramBenchFully Resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 64,000 tokens	0 %	Higher measured value0.5 %	Vals AI Source details Data as of: 08/09/2026
SkillsBenchWith Skills Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Harness: OpenHands, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Harness: OpenHands, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	Higher measured value51.711 %	45.077 %	Vals AI Source details Data as of: 08/07/2026
SWE-bench Verified>4 hours Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	0 %	Higher measured value66.667 %	Vals AI Source details Data as of: 08/08/2026
TaxEval v2Stepwise Reasoning Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	82.993 %	Higher measured value84.628 %	Vals AI Source details Data as of: 08/09/2026
Code MigrationCode Quality Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: mean reviewer code-quality ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	3.632 / 5	Higher measured value3.981 / 5	Vals AI Source details Data as of: 08/09/2026
CorpFin v2Shared Max Context Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value66.783 %	66.317 %	Vals AI Source details Data as of: 08/06/2026
Harvey's Legal Agent BenchmarkArbitration International Dispute Resolution · Criteria passed Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: criteria pass rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	66.667 %	Higher measured value91.756 %	Vals AI Source details Data as of: 08/07/2026
LegalBenchIssue Tasks Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value90.526 %	89.506 %	Vals AI Source details Data as of: 08/09/2026
LiveCodeBench v6Medium Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value85.64 %	74.151 %	Vals AI Source details Data as of: 08/09/2026
LMArena Text, Style ControlBusiness, management and finance Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-high, Reasoning: high, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 12,346 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 6,609 anonymous pairwise votes.	Higher measured value1,483.779 95% confidence interval: ±6.654The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	1,460.145 95% confidence interval: ±8.046The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProChemistry Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	91.078 %	Higher measured value91.608 %	Vals AI Source details Data as of: 08/09/2026
ProgramBenchRaw Pass Rate Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 64,000 tokens	Higher measured value68.596 %	62.605 %	Vals AI Source details Data as of: 08/09/2026
SWE-bench Verified1-4 hours Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	50 %	Higher measured value66.667 %	Vals AI Source details Data as of: 08/08/2026
Harvey's Legal Agent BenchmarkArbitration International Dispute Resolution · Task fully resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 08/07/2026
LegalBenchRhetoric Tasks Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value82.824 %	81.263 %	Vals AI Source details Data as of: 08/09/2026
LMArena Text, Style ControlChinese Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 3,684 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 2,084 anonymous pairwise votes.	1,494.074 95% confidence interval: ±10.702The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	Higher measured value1,516.828 95% confidence interval: ±13.481The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProComputer Science Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	89.756 %	Higher measured value90.244 %	Vals AI Source details Data as of: 08/09/2026
SWE-bench Verified15 min - 1 hour Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	76.245 %	Higher measured value81.609 %	Vals AI Source details Data as of: 08/08/2026
Harvey's Legal Agent BenchmarkBanking Finance · Criteria passed Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: criteria pass rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	80.297 %	Higher measured value85.874 %	Vals AI Source details Data as of: 08/07/2026
LegalBenchRule Tasks Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value84.008 %	78.84 %	Vals AI Source details Data as of: 08/09/2026
LMArena Text, Style ControlChinese Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-high, Reasoning: high, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 3,482 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 2,084 anonymous pairwise votes.	1,504.962 95% confidence interval: ±10.923The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	Higher measured value1,516.828 95% confidence interval: ±13.481The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProEconomics Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	89.692 %	Higher measured value90.403 %	Vals AI Source details Data as of: 08/09/2026
Harvey's Legal Agent BenchmarkBanking Finance · Task fully resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 08/07/2026
LMArena Text, Style ControlCoding Test details Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4, Harness: Arena text, style control GLM-5.2: Snapshot: glm-5.2-max, Reasoning: max, Harness: Arena text, style controlGPT-5.4: Style-controlled Arena rating from 17,704 anonymous pairwise votes.GLM-5.2: Style-controlled Arena rating from 9,517 anonymous pairwise votes.	Higher measured value1,513.577 95% confidence interval: ±5.956The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	1,509.051 95% confidence interval: ±7.106The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.	LMArena Source details Data as of: 09/01/2026
MMLU-ProEngineering Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens	Higher measured value85.243 %	83.385 %	Vals AI Source details Data as of: 08/09/2026
Harvey's Legal Agent BenchmarkBankruptcy Restructuring · Criteria passed Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: criteria pass rateScoring: Higher is betterGPT-5.4: Snapshot: gpt-5.4-2026-03-05, Reasoning: xhigh, Output limit: 128,000 tokens GLM-5.2: Reasoning: max, Temperature: 1, Top P: 0.95, Output limit: 131,072 tokens	68.512 %	Higher measured value74.394 %	Vals AI Source details Data as of: 08/07/2026
Key to the marks:better valuebetter, but not conclusivetie

How to read this comparison

Winning one benchmark is not an overall verdict

Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.

The standard price table uses direct developer rates. The provider comparison labels OpenRouter endpoints, regions, and special tariffs separately.

Every result links to its measurement source and retrieval date. Labels distinguish vendor reports, official benchmark leaderboards, and independent evaluations. Disputed or archived results are excluded.

More matchups

Compare GPT-5.4 and GLM-5.2 with other leading models using the same data and benchmark logic.

View the complete LLM comparison