Back to Models
Zhipu AIOpen Source

GLM-4.5V

GLM-4.5V is a multi-modal foundation model developed by Zhipu AI. It features a large-scale architecture with approximately 1.06 trillion parameters and supports a 64K context window.

Parameters

1060.0B

Context Window

64K

License

MIT

Release Date

2025-08-11

API Pricing

API pricing for this model is not yet available

Strengths

  • Extremely large parameter count exceeding 1 trillion
  • Flexible 64K context length
  • Open usage under MIT license

Weaknesses

  • Massive model size of 312GB
  • Requires very high computational resources to run
  • High memory consumption during inference

Use Cases

  • Advanced multimodal understanding
  • Large-scale dataset analysis
  • Open-source based AI development

Deep Analysis

Parameters

106B total / 12B active MoE

Context Window

32K tokens (128K in successor)

Modalities

Video, Image, Text, File

Pricing

$0.60/$1.80 per 1M tokens

License

Apache 2.0

Release Date

July 2025

Strengths

  • SOTA among open-source VLMs (42 benchmarks)
  • Competitive with Gemini-2.5-Flash on coding
  • Thinking/non-thinking modes
  • Apache 2.0

Weaknesses

  • 16K max output limiting
  • 32K training context
  • Superseded by GLM-4.6V

Competitor Comparison

ModelPrice
Gemini-2.5-FlashN/A
Qwen2.5-VL-72BN/A
GLM-4.6V$0.30/$0.90

GLM-4.5V is Zhipu's open-source VLM with 106B/12B active params. SOTA across 42 visual benchmarks. Apache 2.0.

Sources

Analysis generated: 2026-05-24