Atlas

Models

← All models

GLM-5.3

Zhipu AI (Z.ai)

GLM-5.3 is Z.ai's open-weight post-training update to the GLM-5.2 base model, focused on complex coding and long-horizon tasks. Its reasoning effort supports low, high, and max, with max as the documented default.

Benchmark scores

BenchmarkScore
AA-Briefcase Elo1505
AA-Briefcase v1.1 Elo1525
AA-LCR v1.179.7
AA-Omniscience Index14.3
Agent Red Teaming k=100 (CAIS)69.2
ALE-Bench1937
Antidote: Everyday Edition1067
Arena Code WebDev — Brand & Marketing1636
Arena Code WebDev — Consumer Product1566
Arena Code WebDev — Content Creation Tools1593
Arena Code WebDev — Data & Analytics1560
Arena Code WebDev — Gaming1651
Arena Code WebDev — HTML1579
Arena Code WebDev — React1629
Arena Code WebDev — Reference-Based Design1635
Arena Code WebDev — Simulations1633
Arena Text — Business, Management & Financial Ops (No Style Control)1455
Arena Text — Chinese (No Style Control)1516
Arena Text — Creative Writing (No Style Control)1462
Arena Text — English (No Style Control)1489
Arena Text — Entertainment, Sports & Media (No Style Control)1453
Arena Text — Exclude Ties (No Style Control)1481
Arena Text — Exclude Ties (Style Controlled)1493
Arena Text — Expert (No Style Control)1517
Arena Text — French (No Style Control)1501
Arena Text — French (Style Controlled)1519
Arena Text — German (No Style Control)1528
Arena Text — Hard Prompts (No Style Control)1491
Arena Text — Hard Prompts English (No Style Control)1505
Arena Text — Instruction Following (No Style Control)1481
Loading Atlas data…