glm-5.3

1,613 8 hours ago

Z.ai's flagship model and the most capable open-weights model for coding, with major gains on long-horizon agentic tasks.

tools thinking cloud
Usage
high
Context
1M tokens
Size
753B parameters 
ollama run glm-5.3:cloud

Applications

Claude Code
Claude Code ollama launch claude --model glm-5.3:cloud
OpenCode
OpenCode ollama launch opencode --model glm-5.3:cloud
Hermes Agent
Hermes Agent ollama launch hermes --model glm-5.3:cloud
OpenClaw
OpenClaw ollama launch openclaw --model glm-5.3:cloud

Models

View all →

1 model

glm-5.3:cloud

High Usage · 1M context window · Text · 8 hours ago

Readme

This model is hosted in the United States and Europe. Like all other models on Ollama’s cloud, this model follows Ollama’s privacy policy with zero data retention.

GLM-5.3 is Z.ai’s flagship model and the most capable open-weights model for coding, with major gains on long-horizon agentic tasks.

GLM-5.3 uses the same base model as GLM-5.2 — every gain comes from post-training. Compared with GLM-5.2, it is much better at complex coding and long-horizon tasks, including a 50% improvement on Z.ai’s in-house Code Bench.

What’s new

  • Stronger coding: Terminal-Bench 3.0 improves from 4.6 to 28.3, DeepSWE v1.1 from 46.2 to 66.9, and Terminal-Bench 2.1 from 81.0 to 88.2 over GLM-5.2.
  • Long-horizon agentic tasks: SWE-Marathon more than doubles (19.4 → 42.5), with the top scores in Z.ai’s evaluations on AutomationBench (48.2) and GDPval-AA v2 (1769).
  • Security research: state of the art on CyberGym (84.5) for vulnerability discovery.
  • Adjustable reasoning effort: reasoning_effort accepts low, high, and max, and defaults to max. For chat, set clear_thinking to true.
  • 1M token context window: the same long context as GLM-5.2, for project-level engineering work and long-running sessions.

image.png

Agentic coding

GLM-5.3 is built for coding agents, including Claude Code, OpenCode, Hermes Agent, and OpenClaw. Z.ai’s results below were measured in the Claude Code harness with maximum reasoning effort.

Benchmarks

Benchmark GLM-5.3 GLM-5.2 Kimi K3 DeepSeek-V4 Pro-0813 Qwen3.8-Max Opus 4.8 Fable 5 GPT-5.6 Sol
Terminal Bench 2.1 88.2 81.0 88.3 87.9 86.6 85.0 88.0 88.8
Terminal Bench 3.0 28.3 4.6 17.4 – – 21.1 33.7 34.6
DeepSWE (v1.1) 66.9 46.2 67.5 62.7 56.6 58.0 69.7 72.7
NL2Repo 58.0 48.9 58.0 61.1 55.9 69.7 – –
ProgramBench (Almost Solved) 19.0 9.5 17.5 – 10.5 15.5 33.0 23.0
FrontierSWE 78.1 67.5 – – – 66.5 88.2 –
SWE-Marathon (v1.1) 42.5 19.4 48.1 – – 48.8 33.1 42.5
PostTrainBench 39.8 31.7 32.0 – – 32.9 41.8 36.2
CyberGym 84.5 77.2 80.0 83.3 78.5 78.1 83.8 83.6
ExploitGym (2h / 6h) 105 / 130 29 / 39 36 / 70 – 14 / 26 80 / 120 181 / 247 216 / 293
ExploitBench 54.4 24.4 32.2 – 28.8 40.0 78.0 76.5
Toolathlon Verified 73.0 59.9 76.5 74.1 72.5 76.2 74.7 74.9
AutomationBench (v1.0.6) 48.2 26.2 46.7 43.2 39.8 41.0 46.2 45.8
Agents’ Last Exam (ALE-CLI) 28.5 23.8 27.6 25.7 27.0 25.7 23.8 28.6
HLE w/ Tools 62.5 54.7 59.8 60.0 56.2 57.9 63.9 64.5
GDPval-AA v2 1769 1508 1682 1590 1739 1588 1743 1730

Reference