GLM 5.3 Flash

ox-alpha out of stealth! GLM-5.3 Flash is Z.ai's first natively multimodal GLM-5 model, with 320B total parameters and just 18B active parameters for efficient coding, agentic work, and precise 1M-token context. Its hybrid sparse-and-linear attention architecture helps it outperform GLM-5.2 at one-tenth the price while approaching Claude Opus 4.8 on coding and agentic benchmarks.

Reasoning
Vision
Video Input
Tool Calling
Structured Output

Added Aug 26, 2026

Model weights

Pricing

Auto routing · per 1M tokens

Input: $0.10
Output: $0.30
Cache read: $0.025

Compare provider prices

Specifications

Context window: 1M
Max output: 131.1K
Parameters: 320B / 18B; Total / active
Avg output (7d): 1.3K tokens; Longer than 80% of models

Benchmarks

Artificial Analysis

LMArena

Sourced from Artificial Analysis.

Intelligence Index

41.8

Better than 94% of models compared

Coding Index

71.5

Better than 85% of models compared

Agentic Index

50.9

Better than 92% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

60.4%

Better than 86% of models compared

AutomationBench-AA Tasks Completed

Fully completed workflows without guardrail violations

25.4%

Better than 35% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

1454 Elo

Better than 81% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1647 Elo

Better than 94% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

15.4%

Better than 58% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

80.0%

Better than 87% of models compared

MLCR-AA

Medical long-context reasoning

51.1%

Better than 88% of models compared

Reasoning

HLE

Humanity's Last Exam

39.9%

Better than 88% of models compared

CritPt

Research-level physics reasoning

15.4%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

32.8%

Better than 82% of models compared

SciCode

Python programming for scientific computing

51.6%

Better than 61% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

27.5%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

27.6%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

91.2%

Better than 92% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

80.0%

Better than 87% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

57.3%

Last updated Oct 5, 2026