HolyGrAIL

All tasks / Models

GLM-5.3-Flash

Z.ai's smaller open-weights GLM-5.3 MoE (about 320B total, 18B active) that accepts text, image and video input.

  • Z.ai
  • Open weights
  • Released 26 Aug 2026

Official page

Where it ranks

  • A
    Chat & assistants
    Models · Local · 1,474 on LMArena Text Arena (overall, style control)
  • A
    Writing
    Models · Local · 1,434 on LMArena Text Arena: Creative Writing (style control)
  • –
    Documents & OCR
    Models · Local · no result on this task’s benchmarks yet
  • C
    Coding
    Models · Local · 1,607 on LMArena WebDev / Code Arena (overall)
  • S
    Image understanding
    Models · Local · 1,273 on LMArena Vision Arena (overall, style control)
  • C
    Vector graphics & logos
    Models · Local · 1,286 on Design Arena SVG
  • B
    Workflow automation
    Models · Local · 60.4% on AutomationBench-AA (Artificial Analysis)

Facts

Vendor
Z.ai
Family
GLM Flash
Released
26 Aug 2026
Weights
Open: download and run it yourself
License
MIT
Parameters
321.3B (18B active)
GPU memory to run locally
~185 GB (320B total / 18B active as reported (HF safetensors total 321.3B); 4-bit ≈ 321.3 B × 0.57 GB + 1.5 GB ≈ 185 GB, estimate.)
API price
$0.15 in / $0.5 out per 1M tokens (Z.ai first-party endpoint (fp8) on OpenRouter; third-party hosts list from $0.048/$0.20.)
Access
Free (open weights)
Last verified
18 Oct 2026

Price from openrouter.ai, 18 Oct 2026.

Benchmark results

BenchmarkResultEntry on the leaderboardAs ofSource
Artificial Analysis Intelligence Index42GLM-5.3-Flash
Raw value from the page's embedded model data; the site displays it rounded to an integer.
19 Oct 2026artificialanalysis.ai
AutomationBench-AA (Artificial Analysis)60.4%GLM-5.3-Flash
AutomationBench-AA Score (share of each task's objectives completed, a task with any guardrail violation scoring zero); raw value from the page's embedded model data (displayed rounded); rank 17 of 32; Tasks Completed 25.42%, Objectives Completed 86.45%; no last-updated date shown, so as_of = retrieval date.
19 Oct 2026artificialanalysis.ai
Design Arena SVG1,286GLM-5.3-Flash
Design Arena SVG, rank 16, 2102 battles; board id 'notebook'; rows read from POST https://www.designarena.ai/api/leaderboard (arenaType 'models', category 'svg'); names from https://www.designarena.ai/api/registry
19 Oct 2026designarena.ai
LMArena Text Arena (overall, style control)1,474glm-5.3-flash
From LMArena's official leaderboard-dataset (text_style_control, split latest, category overall, published 2026-10-15), which mirrors https://arena.ai/leaderboard/text (site blocks automated fetches). Rank 41 of 416; 32925 votes; 95% CI 1469.0-1478.7.
15 Oct 2026huggingface.co
LMArena Text Arena: Creative Writing (style control)1,434glm-5.3-flash
From LMArena's official leaderboard-dataset (text_style_control, split latest, category creative_writing, published 2026-10-15), which mirrors https://arena.ai/leaderboard/text/creative-writing (site blocks automated fetches). Rank 68 of 414; 6983 votes; 95% CI 1425.5-1441.8.
15 Oct 2026huggingface.co
LMArena Vision Arena (overall, style control)1,273glm-5.3-flash
From LMArena's official leaderboard-dataset (vision_style_control, split latest, category overall, published 2026-10-09), which mirrors https://arena.ai/leaderboard/vision (site blocks automated fetches). Rank 35 of 165; 11157 votes; 95% CI 1266.3-1280.4.
9 Oct 2026huggingface.co
LMArena WebDev / Code Arena (overall)1,607GLM-5.3-Flash
From LMArena's official leaderboard-dataset (webdev = Code Arena, split latest, category overall, published 2026-10-18), which mirrors https://arena.ai/leaderboard/webdev. Rank 28; 18002 votes; 95% CI 1600.3-1613.4.
18 Oct 2026huggingface.co
ParseBench (models)70.8%GLM-5.3 Flash
Overall (mean of tables, charts, content faithfulness, semantic formatting, visual grounding) from leaderboard.csv in the ParseBench repo (last leaderboard commit 2026-10-09).
9 Oct 2026github.com

Where you can use it

Compare

Alternatives to GLM-5.3-Flash

Sources