GLM-5.2
Active Parameters
744B
Context Length
1M
Modality
Text
Architecture
Mixture of Experts (MoE)
License
MIT
Release Date
13 Jun 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Input: $1.40 · Output: $4.40
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
98x RTX 4090
24GB VRAM
Datacenter
25x NVIDIA A100
80GB VRAM
Apple Silicon
21x Apple M3 Max
128GB VRAM
1,000,000 tokens
Consumer
491x RTX 4090
24GB VRAM
Datacenter
106x NVIDIA A100
80GB VRAM
Apple Silicon
113x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
Rank
#20
| Benchmark | Score | Rank |
|---|---|---|
StackUnseen | 0.867 | 9 |
Coding | 0.80 | 17 |
Agent Arena | max 0.04 | 18 |
Graduate-Level QA | 0.912 | 20 |
Web Development | max 1598 | 21 |
Agentic Coding | 0.52 | 30 |
Mathematics | 0.90 | 30 |
Data Analysis | 0.74 | 34 |
General Text | max 1472 | 37 |
General | 0.73 | 39 |
LiveBench Average | 0.73 | 39 |
Reasoning | 0.79 | 43 |
Agentic Index | max 0.38 | 45 |
Coding Index | max 0.69 Standard 0.47 | 50 98 |
Intelligence Index | max 0.34 Standard 0.22 | 69 138 |
Rankings
Overall Rank
#20
Coding Rank
#14
About GLM-5.2
Z.ai's flagship open-weights foundation model released June 13, 2026. A 744-billion-parameter Mixture-of-Experts model with 40 billion active parameters per token, powered by IndexShare technology. Features a stable 1M-token context window optimised for long-horizon coding and agentic workflows, multiple thinking-effort levels (High and Max), and strong benchmark performance (81.0 on Terminal-Bench 2.1, 62.1 on SWE-bench Pro). Supports thinking mode, streaming, function calling, context caching, structured output, and MCP integration. Released under MIT license with no regional restrictions. Priced at $1.40 per million input tokens and $4.40 per million output tokens.
Technical Specifications
Attention
Attention Structure
DeepSeek Sparse Attention
Attention Heads
64
Key-Value Heads
64
Attention Head Dimension
192
Position Embedding
ROPE
RoPE Theta
8,000,000
Sliding Window Attention
No
Sliding Window Size
-
Sliding Window Ratio
-
Linear Attention
-
Linear Attention Ratio
-
Normalization
RMS Normalization
Activation Function
Swish
Dimensions
Auxiliary Parameters
-
Hidden Dimension Size
6,144
Number of Layers
78
FFN Intermediate Size (Dense)
12,288
Multi-Token Prediction Heads
1
Tokenizer
Vocabulary Size
154,880
Mixture of Experts
Total Expert Parameters
40.0B
Number of Experts
256
Active Experts
8
Shared Experts
1
FFN Intermediate Size (per Expert)
2,048
Dense Layers Before MoE
3
About GLM-5.2
Z.ai's GLM-5.2 is a 744-billion-parameter Mixture-of-Experts flagship foundation model, released June 13, 2026, designed for long-horizon coding and agentic engineering tasks. It features a usable 1M-token context window, IndexShare architecture reducing compute to 1/20th of prior generations, and multiple thinking-effort levels. Open-weights under MIT license with no regional restrictions.
Other GLM-5.2 Models
- No related models available