GLM-4.5-MoE-355B-A32B-0715
GLM-4.5-MoE-355B-A32B-0715 is a reasoning model developed by Zhipu AI. Equipped with a massive parameter scale of 3550B and a 128K context window, it provides advanced reasoning capabilities.
Parameters
3550.0B
Context Window
128K
License
Apache 2.0
Release Date
2025-07-28
API Pricing
API pricing for this model is not yet available
Strengths
- ・Overwhelming 3550B parameter count
- ・Supports long 128K context
- ・Open-source Apache 2.0 license
Weaknesses
- ・Extremely large 710GB file size
- ・High operational load from massive resources
- ・Increased memory use due to MoE structure
Use Cases
- ・Reasoning tasks requiring complex logical thought
- ・Long-form analysis handling massive documents
- ・Advanced open-source AI development
Deep Analysis
Architecture
MoE (355B total, 32B active)
Flagship GLM-4.5 model
Context Window
128K tokens
Maximum Output
96K tokens
Release Date
July 2025
License
Open weights
Training Data
15T tokens
General + code/reasoning/agent fine-tuning
Strengths
- ・Most powerful GLM-4.5 model
- ・32B active from 355B total for efficiency
- ・96K max output tokens
- ・Supports Claude Code and Roo Code integration
- ・Hybrid reasoning (Thinking/Non-Thinking modes)
- ・Strong on AIME, GPQA, SWE-Bench
Weaknesses
- ・Large total parameter count limits self-hosting
- ・Chinese model origin
- ・Context shorter than some competitors (128K vs 200K+)
Competitor Comparison
| Model | Arena | SWE | GPQA | Price |
|---|---|---|---|---|
| Claude Sonnet 4.6 | - | - | - | Higher |
| DeepSeek V3.2 | - | - | - | Lower |
| Kimi K2.5 | - | - | - | Comparable |
GLM-4.5 (355B-A32B) is Zhipu AI's flagship open-weight reasoning model with 355B total parameters and 32B active. It combines advanced reasoning, coding, and agent capabilities with native tool invocation support, achieving performance comparable to Claude and DeepSeek.
Analysis generated: 2026-05-24