AI Model Reviewer

GLM-5.3

Z.ai. Run in Claude Code, OpenCode, Raw API call. Served via Fireworks; GLM has no coding CLI of its own, so its home track runs in Claude Code (Z.ai's recommended harness), labelled as such.

110 runs on 77 prompts · 41% finished

Results

The best run of each of its 77 tests first, then every other run. Filter them in the explorer

45 finished runs

  • GLM-5.3, OpenCode: Isometric SVG town with sprite sheet

    Isometric SVG town with sprite sheet

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, OpenCode: Pelican riding a bicycle (control)

    Pelican riding a bicycle (control)

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, OpenCode: Animated pelican riding a bicycle (SMIL/CSS)

    Animated pelican riding a bicycle (SMIL/CSS)

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, OpenCode: MacBook Pro 16" realistic 3D render (Playcode benchmark)

    MacBook Pro 16" realistic 3D render (Playcode benchmark)

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, OpenCode: Lighthouse on a headland at night

    Lighthouse on a headland at night

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, Claude Code: Fjord scene (Three.js)Live scene

    Fjord scene (Three.js)

    GLM-5.3

    Claude Code 2.1.280

    skills off

  • GLM-5.3, OpenCode: Video printing press collage remotionRendered video

    Video printing press collage remotion

    GLM-5.3

    OpenCode 1.18.32

    skills off

    Rendered video from the run

  • GLM-5.3, OpenCode: Spongebob

    Spongebob

    GLM-5.3

    OpenCode 1.18.32

    skills off

    3D model exported by the site · Explore 3D model

  • GLM-5.3, Raw API (one-shot): Desert Dunes at Night

    Desert Dunes at Night

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Sales Dashboard Card

    Sales Dashboard Card

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Calendar Month View

    Calendar Month View

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Exploded Burger

    Exploded Burger

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Bauhaus Composition

    Bauhaus Composition

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Basketball Slam Dunk

    Basketball Slam Dunk

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Battery Level Icons

    Battery Level Icons

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Fruit Sales Bar Chart

    Fruit Sales Bar Chart

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Bohr Model of an Atom

    Bohr Model of an Atom

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Flat App Icon Set

    Flat App Icon Set

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Alien Planet

    Alien Planet

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Acoustic Guitar

    Acoustic Guitar

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Raw API (one-shot): Sleeping Curled Cat

    Sleeping Curled Cat

    GLM-5.3

    Raw API (one-shot)

    raw API baseline

  • GLM-5.3, Claude Code: GPU fluid simLive scene

    GPU fluid sim

    GLM-5.3

    Claude Code 2.1.280

    skills off

  • GLM-5.3, OpenCode: Pelican car france

    Pelican car france

    GLM-5.3

    OpenCode 1.18.32

    skills off

  • GLM-5.3, OpenCode: Dolphin vlog bicycle

    Dolphin vlog bicycle

    GLM-5.3

    OpenCode 1.18.32

    skills off

Numbers: finish rate, cost, wall time, draft scores
Runs
110
on 77 prompts
Finished
41%
45 of 110
Median cost, finished run
$1.08
n=31
Median wall time, finished run
17m 09s
n=45
Recorded spend
$73.00
$20.54 on unfinished runs; 61 without a recorded cost

From the numbers

Set against the other models with at least 3 runs (top or bottom third), and its draft-score ranks within single prompts. Nothing here is averaged across prompts.

Strengths

  • Draft score 9.5/10 on Pelican riding a bicycle (control) (in OpenCode): ranked 1st of 13
  • Draft score 9/10 on MacBook Pro 16" realistic 3D render (Playcode benchmark) (in OpenCode): ranked 2nd of 12
  • Draft score 9.5/10 on Isometric SVG town with sprite sheet (in OpenCode): ranked 2nd of 11

Weaknesses

  • 45 of 110 runs finished: 22nd of 24 models with at least 3 (finish rate)
  • 17m 09s (n=45): 20th of 23 models with at least 3 (median wall time of a finished run)
  • Draft score 5.5/10 on Lighthouse on a headland at night (in OpenCode): ranked 12th of 12
  • 44 runs ended as the model's own failure (DNF, class MODEL)
  • Other unfinished runs, not counted against the model: 9 HARNESS, 4 INFRA, 7 TIMEOUT, 1 ERROR

Best and hardest run

Best run

Picked as its best draft-score rank on a prompt.

Hardest run

Picked as its costliest run that did not finish.

Draft scores by prompt

Isometric SVG town with sprite sheetOpenCode9.5/10 draft2nd of 11
Lighthouse on a headland at nightOpenCode5.5/10 draft12th of 12
MacBook Pro 16" realistic 3D render (Playcode benchmark)OpenCode9/10 draft2nd of 12
Pelican riding a bicycle (control)OpenCode9.5/10 draft1st of 13
Animated pelican riding a bicycle (SMIL/CSS)OpenCode9/10 draft6th of 12

By harness

OpenCode241771%$1.08 n=1726m 23s n=17
Claude Code291448%$1.07 n=1421m 34s n=14
Raw API call571425%not recorded: no finished run with a recorded cost11m 22s n=14
All 77 tests and experiment collections