Z.ai

GLM-5.1

Name: GLM-5.1
Author: Z.ai

Frontier

HuggingFace

141.5KDownloads1.7KCurtidasApr 2026Publicado200K tokensContextoMITLicença92 ExcepcionalQualidade

GLM-5.1 (754B parameters) requires approximately 482.0 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 555 GB of VRAM.

Comece agora

— copie e cole para rodar localmente

Copy-paste commands to run GLM-5.1 on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "zai-org/GLM-5.1" \
  --hf-file "GLM-5.1-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Quick specs

Parameters754B (40B active)

Architecturemoe (MoE)

Context200K tokens

Modalitytext

Min RAM294.1 GB

Rec. RAM459.9 GB (Q4_K_M)

LicenseMIT

FamilyGLM

✓ Code✓ Chat✓ Reasoning

About this model

GLM-5.1 is Z.ai's next-generation flagship MoE model for agentic engineering, with significantly stronger coding capabilities than GLM-5. It achieves state-of-the-art performance on SWE-Bench Pro and sustains optimization over hundreds of rounds and thousands of tool calls on long-horizon agentic tasks.

•Agentic engineering focus: leads GLM-5 by a wide margin on NL2Repo (repo generation) and Terminal-Bench 2.0 (real-world terminal tasks).
•State-of-the-art SWE-Bench Pro performance (58.4), surpassing GLM-5, Claude Opus 4.6, and GPT-5.4.
•Built to stay effective over much longer horizons — breaks complex problems down, runs experiments, reads results, and revises strategy through repeated iteration.
•Uses DeepSeek Sparse Attention (DSA) MoE architecture (256 routed experts, 8 active per token, 1 shared) for reduced deployment cost.

Modelos relacionados

Opções de quantização

Estimativas de VRAM por nível de quantização

No hardware detected — fit column shows raw VRAM estimates

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	294.1 GB	Low	—
Q3_K_S	3	369.5 GB	Low	—
NVFP4	4	422.2 GB	Medium	—
Q4_K_M	4	459.9 GB	Medium	—
Q5_K_M	5	542.9 GB	High	—
Q6_K	6	618.3 GB	High	—
Q8_0	8	806.8 GB	Very High	—
F16	16	1545.7 GB	Maximum	—

Quality benchmarks

GLM-5.1 benchmark scores

Benchmark verified

Reasoning

MMLU-Pro—

GPQA Diamond86.2%

MATH-500—

ARC Challenge—

Source: official · 2026-04-03

Compatibilidade de hardware

Estimativas de compatibilidade para todo o hardware

Abrir calculadora

Computing compatibility...

Detalhamento de memória

Reference: RTX 2060 6GB

Weights459.9 GB

KV Cache19.0 GB

Runtime2.4 GB

Headroom0.6 GB

Perguntas frequentes

FAQ — GLM-5.1

Veja também

Guia de Quantização Metodologia de Pontuação Abrir calculadora