Z.ai
GLM-5
前沿
120.8K下载量2.1K点赞Feb 2026发布日期200K tokens上下文Custom许可证91 卓越质量
GLM-5 (744B parameters) requires approximately 475.9 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 40B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 548 GB of VRAM.
快速开始
— 复制粘贴即可本地运行Copy-paste commands to run GLM-5 on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "zai-org/GLM-5" \
--hf-file "GLM-5-Q4_K_M.gguf" \
-c 4096 -ngl 99Quick specs
Parameters744B (40B active)
Architecturemoe (MoE)
Context200K tokens
Modalitytext
Min RAM290.2 GB
Rec. RAM453.8 GB (Q4_K_M)
LicenseCustom
FamilyGLM
✓ Code✓ Chat✓ Reasoning
About this model
- •Humanity’s Last Exam (HLE) & other reasoning tasks: We evaluate with a maximum generation length of 131,072 tokens (temperature=1.0, top_p=0.95,...
- •SWE-bench & SWE-bench Multilingual: We run the SWE-bench suite with OpenHands using a tailored instruction prompt. Settings: temperature=0.7,...
- •BrowserComp: Without context management, we retain details from the most recent 5 turns. With context management, we use the same discard-all...
- •Terminal-Bench 2.0 (Terminus 2): We evaluate with the Terminus framework using timeout=2h, temperature=0.7, top_p=1.0, max_new_tokens=8192, with a...
- •Terminal-Bench 2.0 (Claude Code): We evaluate in Claude Code 2.1.14 (think mode, default effort) with temperature=1.0, top_p=0.95,...
相关模型
量化选项
各量化级别的 VRAM 估算
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 290.2 GB | Low | — |
Q3_K_S | 3 | 364.6 GB | Low | — |
NVFP4 | 4 | 416.6 GB | Medium | — |
Q4_K_M | 4 | 453.8 GB | Medium | — |
Q5_K_M | 5 | 535.7 GB | High | — |
Q6_K | 6 | 610.1 GB | High | — |
Q8_0 | 8 | 796.1 GB | Very High | — |
F16 | 16 | 1525.2 GB | Maximum | — |
Quality benchmarks
GLM-5 benchmark scores
Coding
SWE-bench Verified77.8%
HumanEval+—
Aider Polyglot—
LiveCodeBench—
Reasoning
MMLU-Pro70.4%
GPQA Diamond86.0%
MATH-500—
ARC Challenge—
Source: official · 2026-02-20
硬件兼容性
全部硬件的适配估算
Computing compatibility...
内存详细分析
Reference: RTX 2060 6GB
Weights453.8 GB
KV Cache19.0 GB
Runtime2.4 GB
Headroom0.6 GB
常见问题
FAQ — GLM-5
另请参阅