The LLM benchmark aggregator.

GLM-5.3 Flash

Z.ai · 321B · open weights

Strongest in Document parsing (#16 of 83), weakest in Jailbreak resistance (#230 of 268). Above par in 22 of 34 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5.1 and Claude Fable 5 and ahead of Qwen3 235B A22B Instruct and Kimi K2.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Knowledge57.5−5#18/1972/3

Document parsing58−5.3#16/831/1

Academic knowledge56−7.9#48/1231/1

Human preference63.8−5#33/3541/1

Professional53.6−8.8#52/1942/2

Document design51−17.7#17/311/1

Finance57.6−4.8#23/1621/1

Medical57.5−9#25/1491/1

Cybersecurity52.6−11.3#27/471/1

IT operations39.6−34.2#29/331/1

Legal54.5−11.4#50/1731/1

Coding51.2−15#80/1794/6

Agentic coding51.2−15.7#78/1734/5

Agents52.9−13.5#83/2983/6

Long-horizon work40.9−33.1#22/311/1

Web research46.4−17.5#26/351/2

Knowledge work64.3−6.4#27/2082/2

Computer use50.7−17.4#33/681/2

Core abilities45.3−20#142/1993/4

Data analysis53.8−11.4#18/311/1

Language49−17.6#38/621/1

Instruction following26−47.8#62/621/1

General intelligence48.9−20.5#100/1782/3

Reasoning43.8−21.6#149/1954/5

Science57.2−6.3#43/1221/1

Reasoning45.1−23.9#88/1472/3

Mathematics36.7−27.1#132/1522/2

Safety50.2−11.2#173/3291/3

Fairness66.4−3.6#17/2921/2

Toxicity avoidance55.5−3.5#91/2681/1

Secure code51−15.2#148/2701/1

Harm refusal46.2−15.3#218/2921/2

Jailbreak resistance35.2−29.8#230/2681/1

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for GLM-5.3 Flash, left for the other.

Gemini 3.5 Flash-Litewon 21 · lost 1

Command Awon 20 · lost 2

Gemma 4 31Bwon 23 · lost 3

DeepSeek V3won 20 · lost 4

Llama 4 Scout Instructwon 18 · lost 4

Gemini 2.0 Flashwon 18 · lost 4

GPT-5.4 nanowon 18 · lost 4

GPT-4.1 nanowon 19 · lost 5

GPT-4o miniwon 18 · lost 5

GPT-4owon 17 · lost 5

NVIDIA Nemotron 3 Ultra 550B A55Bwon 20 · lost 6

Grok 4.1 Fastwon 16 · lost 5

MiniMax M3won 20 · lost 7

GLM-4.6won 17 · lost 6

Grok 3won 16 · lost 6

MiMo V2.5 Prowon 15 · lost 6

Llama 4 Maverick Instructwon 17 · lost 7

GPT OSS 120Bwon 17 · lost 7

GPT-4.1 miniwon 17 · lost 7

DeepSeek V3.2won 17 · lost 7

Claude Haiku 4.5won 16 · lost 7

DeepSeek R1won 16 · lost 7

Inklingwon 22 · lost 10

GPT OSS 20Bwon 14 · lost 7

Grok 4.3won 16 · lost 8

Gemini 2.5 Prowon 16 · lost 8

Qwen3.8 27Bwon 18 · lost 9

Kimi K2.5won 16 · lost 9

Kimi K2.6won 16 · lost 9

Qwen3.7 Maxwon 16 · lost 9

O3 Miniwon 14 · lost 8

GLM-5.1won 14 · lost 8

GPT-5 miniwon 15 · lost 10

Claude 3.7 Sonnetwon 13 · lost 9

GPT-5 nanowon 12 · lost 9

MiniMax M2.7won 12 · lost 9

Mistral Large 4won 16 · lost 12

Gemini 2.5 Flashwon 13 · lost 10

Grok 4won 13 · lost 10

Claude Sonnet 4won 14 · lost 11

Gemini 3.6 Flashwon 15 · lost 12

MiMo V2.6 Flashwon 12 · lost 10

GPT-6 Lunawon 17 · lost 15

O4 Miniwon 12 · lost 11

Kimi K2won 12 · lost 11

Qwen3 235B A22B Instructwon 12 · lost 11

O3won 12 · lost 12

Claude Sonnet 4.6won 12 · lost 12

GLM-5.2won 13 · lost 13

GPT-5.6 Lunawon 14 · lost 15

GPT-4.1won 11 · lost 12

GPT-5.1won 10 · lost 12

DeepSeek V4 Flashwon 12 · lost 15

Gemini 3.5 Flashwon 12 · lost 15

Claude Sonnet 4.5won 11 · lost 14

Gemini 3.1 Prowon 9 · lost 12

Claude Opus 4won 9 · lost 13

DeepSeek V4 Prowon 11 · lost 17

Grok 4.5won 11 · lost 17

GPT-6 Solwon 10 · lost 16

GPT-5won 9 · lost 15

Gemini 3.8 Flashwon 12 · lost 22

Claude Haiku 5.5won 9 · lost 17

Claude Sonnet 5won 9 · lost 19

Grok 4.6won 10 · lost 22

Claude Sonnet 5.5won 8 · lost 18

GPT-5.2won 7 · lost 18

GPT-5.6 Terrawon 9 · lost 25

Claude Opus 4.6won 6 · lost 17

GPT-6 Astrawon 7 · lost 20

GPT-5.4won 7 · lost 20

Qwen3.8 Maxwon 8 · lost 24

MiMo V2.6 Prowon 6 · lost 22

Muse Spark 1.1won 5 · lost 20

DeepSeek V4.1 Flashwon 5 · lost 21

GPT-6.1 Solwon 5 · lost 21

Gemini 3.7 Flashwon 4 · lost 18

GPT-5.5won 5 · lost 23

Claude Opus 4.7won 4 · lost 23

GLM-5.3won 3 · lost 19

Grok 4.7won 4 · lost 26

GPT-5.6 Solwon 4 · lost 26

Claude Opus 5.5won 4 · lost 28

Kimi K3won 3 · lost 23

Claude Opus 4.8won 3 · lost 24

Muse Wonderwon 2 · lost 20

Gemini 4 Argonwon 2 · lost 21

Claude Opus 5won 2 · lost 26

Muse Spark 1.3won 1 · lost 22

Claude Fable 5won 1 · lost 22

Claude Fable 5.1won 0 · lost 28

§ 3 · Sources

Where the numbers come from

11 publications, 48 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-10

LMArena Text 1474

Artificial Analysis ↗1 measureread 2026-11-10

Artificial Analysis Intelligence Index 28

GDPval-AA ↗1 measureread 2026-11-10

GDPval-AA 1649

AA-Briefcase ↗1 measureread 2026-11-10

AA-Briefcase 1450

Terminal-Bench ↗1 measureread 2026-11-10

Terminal-Bench 35.5% (Claude Code)

ARC-AGI-2 ↗1 measureread 2026-11-10

ARC-AGI-2 65.8%

LiveBench ↗8 measuresread 2026-11-10

LiveBench 60.7LiveBench · Agentic 20.7LiveBench · Reasoning 77.6LiveBench · Coding 70.9LiveBench · Browser Use 59.8LiveBench · Data Analysis 76.4LiveBench · Research 73.9LiveBench · Design 73.7

LiveBench 1.0 ↗8 measuresread 2026-11-10

LiveBench 1.0 71.6LiveBench 1.0 · Reasoning 77.6LiveBench 1.0 · Coding 79LiveBench 1.0 · Agentic Coding 56.8LiveBench 1.0 · Mathematics 81.2LiveBench 1.0 · Data Analysis 76.4LiveBench 1.0 · Language 77.3LiveBench 1.0 · Instruction Following 52.8

Vals.ai ↗14 measuresread 2026-11-10

Vals · Legal Research Bench 45.19%Vals · LegalBench 83.93%Vals · Harvey Legal Agent Benchmark 10.42%Vals · Finance Agent 57.85%Vals · TaxEval 75.59%Vals · MedScribe 88.94%Vals · CyberBench 67.86%Vals · SRE Bench 0.76%Vals · SWE-bench Verified 92% (Mini-SWE-agent)Vals · Vibe Code Bench 30.76% (OpenHands)Vals · Code Migration 20.52%Vals · GPQA Diamond 86.36%Vals · MMLU Pro 86.06%Vals · ProofBench 21%

ParseBench ↗6 measuresread 2026-11-10

ParseBench · Overall 70.55%ParseBench · Tables 89.19%ParseBench · Charts 54.59%ParseBench · Content faithfulness 88.18%ParseBench · Semantic formatting 79.02%ParseBench · Visual grounding 41.78%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-11-10

Enkrypt · Jailbreak risk 25.2%Enkrypt · Harmful content risk 2.2%Enkrypt · CBRN risk 47.2%Enkrypt · Toxicity risk 2.4%Enkrypt · Bias risk 52.5%Enkrypt · Insecure code risk 31.1%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=glm-5-3-flash)](https://publicai.io/model-index/m/glm-5-3-flash)