Z.aiGLM 4Glm4vMoeForConditionalGeneration

GLM 4.6V — Hardware Requirements & GPU Compatibility

VisionFunctions

GLM-4.6V is Z.ai's 107.7-billion-parameter mixture-of-experts vision-language model, the larger of two GLM-4.6V variants (alongside a 9B "Flash" model for local, low-latency deployment) built for visual reasoning, document understanding, and multimodal agents rather than text-only chat. It routes to roughly 14.3 billion active parameters per token and is the first in the GLM-V line to add native multimodal function calling, letting it take images, screenshots, or document pages directly as tool inputs and interpret visual tool outputs like charts or rendered pages as part of its reasoning. It also supports interleaved image-text generation and pixel-accurate UI-to-HTML/CSS reconstruction, and needs a multi-GPU server to run even quantized given its size. Context length is 131,072 tokens. It is released under the MIT license, permitting unrestricted commercial and research use, and was published in December 2025.

107.7B params · MoE14B active131K contextMITDec 20255.4K downloads398 likes57.7K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Doesn't fit a 24 GB GPU

Smallest consumer fit → RTX 5090 · IQ2_XXS · 31 GB · ~168 tok/s

Macs from 36 GB (IQ2_XXS · ~46 tok/s)

MIT

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run GLM 4.6V

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Ways to runOther ways

each opens step-by-step

Run in cloud

Runs on A100 80GB (14 GB headroom) · Q4_K_M

Generation speed: ~111 tok/s; generation speed
Cost per 1M output tokens: $2.81; per 1M output tokens

Deployon Vast.aiBest valuefrom $1.12/hr

Compare GPUs →

Run GLM 4.6V

66.0 GB VRAM at 16K context

Works on macOS, Windows and Linux.

Install Ollama (Linux)
$curl -fsSL https://ollama.com/install.sh | sh
Install Ollama (macOS / Windows)
Download and run the app; it installs the ollama command.
Download Ollama ↗
Run Q4_K_M
$ollama run hf.co/unsloth/GLM-4.6V-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=16384 ollama serve.
$/set parameter num_ctx 16384
A larger context needs more memory than the weights alone.
Use the API
Ollama serves its API on http://localhost:11434.

Works on macOS, Windows and Linux.

Install LM Studio
Desktop app for macOS (Apple silicon), Windows and Linux.
lmstudio.ai ↗
Download the model
In the app press Ctrl/Cmd + Shift + M and paste https://huggingface.co/unsloth/GLM-4.6V-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/unsloth/GLM-4.6V-GGUF@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 16384. Leave GPU offload at maximum if it fits.
Start the local server (optional)
$lms server start

Works on macOS, Windows and Linux.

Install llama.cpp (macOS)
MacPorts, Nix and conda-forge also work.
$brew install llama.cpp
Install options ↗
Install llama.cpp (Windows)
$winget install llama.cpp
Install llama.cpp (Linux)
Download the prebuilt llama-*-bin-ubuntu-* archive for your GPU backend (CPU, Vulkan) from the releases page, unpack it and run the binaries from that folder. Homebrew (brew install llama.cpp) and conda-forge also work on Linux.
llama.cpp releases ↗
Start the server
Downloads the model on first run. The chat UI and an OpenAI-compatible API are at http://localhost:8080.
$llama-server -hf unsloth/GLM-4.6V-GGUF:Q4_K_M -c 16384 -ngl 99 --port 8080

Works on macOS, Windows and Linux.

Option A: Docker Model Runner
Docker Desktop: Settings > AI > Enable Docker Model Runner. Docker Engine on Linux: install the docker-model-plugin package.
$docker model run hf.co/unsloth/GLM-4.6V-GGUF:Q4_K_M
Model Runner docs ↗
The quant tag after the repo is how Hugging Face cards write it; if your Docker version rejects it, drop the tag and pick the quant by repo.
Option B: llama.cpp server image (NVIDIA)
Needs the NVIDIA Container Toolkit. Models are cached in the llama-cache volume. Open http://localhost:8080.
$docker run --gpus all -p 8080:8080 -v llama-cache:/root/.cache ghcr.io/ggml-org/llama.cpp:server-cuda -hf unsloth/GLM-4.6V-GGUF:Q4_K_M -c 16384 -ngl 99 --host 0.0.0.0 --port 8080
Image tags (cuda13, vulkan, rocm) ↗
Cache path inside the image is not in the docs; remove the -v flag if the volume complains.

Linux only.

Install vLLM
Linux only; on Windows run it inside WSL2. NVIDIA GPU shown; AMD ROCm, Intel and CPU builds have their own install pages.
$uv venv --python 3.12 --seed
$source .venv/bin/activate
$uv pip install vllm --torch-backend=auto
vLLM installation ↗
Serve the model
OpenAI-compatible API at http://localhost:8000/v1.
$vllm serve zai-org/GLM-4.6V --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~237 GB of GPU memory.

macOS only.

MLX supports only macOS, so follow these steps for that setup.

Install mlx-lm
$pip install mlx-lm
Chat in the terminal
$mlx_lm.chat --model mlx-community/GLM-4.6V-nvfp4
Downloads the model on first run. The MLX repo has its own quantisation; it may not match the quant selected above.
Or start a local server
Listens on http://localhost:8080.
$mlx_lm.server --model mlx-community/GLM-4.6V-nvfp4

Works on macOS, Windows and Linux.

Install Jan
jan.ai ↗
Find the model
Open the Hub tab, paste this model ID in Search, and choose Q4_K_M if offered.
$unsloth/GLM-4.6V-GGUF
Use the local API (optional)
Settings > Local API Server > Start Server. It listens on http://127.0.0.1:1337.

Rent GPUs hourly to run Q4_K_M right there. Pricing comes from partner marketplaces and varies.

Available GPU rentals for Q4_K_M, prioritized by value
GPU	Speed	$/1M tokens	Rent
A100 80GBCheapestCheapest14 GB free	~111 tok/s	$2.81	Deploy on Vast.aifrom $1.12/hr (opens in new tab)	14 GB free · ~111 tok/s · $2.81 / 1M tokens
A100 80GB14 GB free	~111 tok/s	$3.49	Deploy on RunPodfrom $1.39/hr (opens in new tab)	14 GB free · ~111 tok/s · $3.49 / 1M tokens
RTX PRO 6000 (96 GB)30 GB free	~101 tok/s	$3.87	Deploy on Vast.aifrom $1.40/hr (opens in new tab)	96 GB · 30 GB free · ~101 tok/s · $3.87 / 1M tokens
RTX PRO 6000 Max-Q (96 GB)30 GB free	~101 tok/s	$3.87	Deploy on Vast.aifrom $1.40/hr (opens in new tab)	96 GB · 30 GB free · ~101 tok/s · $3.87 / 1M tokens
RTX PRO 6000 Server (96 GB)30 GB free	~92 tok/s	$4.83	Deploy on Vast.aifrom $1.60/hr (opens in new tab)	96 GB · 30 GB free · ~92 tok/s · $4.83 / 1M tokens
RTX PRO 6000 (96 GB)30 GB free	~101 tok/s	$4.67	Deploy on RunPodfrom $1.69/hr (opens in new tab)	96 GB · 30 GB free · ~101 tok/s · $4.67 / 1M tokens
H100 PCIe (80 GB)14 GB free	~111 tok/s	$5.00	Deploy on RunPodfrom $1.99/hr (opens in new tab)	80 GB · 14 GB free · ~111 tok/s · $5.00 / 1M tokens
H100 SXM (80 GB)14 GB free	~154 tok/s	$4.85	Deploy on RunPodfrom $2.69/hr (opens in new tab)	80 GB · 14 GB free · ~154 tok/s · $4.85 / 1M tokens

Compare GPUs →

How Much VRAM Does GLM 4.6V Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
30.1 GB	32 GB	38.2 GB	41.2 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Does not fit: needs 31 GB
32.7 GB	34.7 GB	40.8 GB	40.5 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Does not fit: needs 33.6 GB
34.1 GB	36 GB	42.2 GB	40.9 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Does not fit: needs 35 GB
36.8 GB	38.7 GB	44.9 GB	42.9 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Does not fit: needs 37.7 GB
42.2 GB	44.1 GB	50.3 GB	49.7 GB	3-bit · i-quant — Importance-weighted 3-bit	Does not fit: needs 43.1 GB
43.5 GB	45.4 GB	51.6 GB	43.7 GB	2-bit · small K-quant — 2-bit small K-quant	Does not fit: needs 44.4 GB
44.9 GB	46.8 GB	53 GB	48.2 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Does not fit: needs 45.8 GB
46.2 GB	48.1 GB	54.3 GB	50.7 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Does not fit: needs 47.1 GB
46.2 GB	48.1 GB	54.3 GB	43.9 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Does not fit: needs 47.1 GB
47.5 GB	49.5 GB	55.7 GB	50.9 GB	3-bit small — 3-bit small quantization	Does not fit: needs 48.5 GB
48.9 GB	50.8 GB	57 GB	51.4 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Does not fit: needs 49.8 GB
52.9 GB	54.9 GB	61 GB	55.4 GB	3-bit medium — 3-bit medium quantization	Does not fit: needs 53.8 GB
54.3 GB	56.2 GB	62.4 GB	60.7 GB	4-bit · legacy — 4-bit legacy quantization	Does not fit: needs 55.2 GB
55.6 GB	57.6 GB	63.7 GB	57.6 GB	3-bit large — 3-bit large quantization	Does not fit: needs 56.5 GB
58.3 GB	60.3 GB	66.4 GB	58.3 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Does not fit: needs 59.2 GB
61.0 GB	63 GB	69.1 GB	67.1 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Does not fit: needs 61.9 GB
61.0 GB	63 GB	69.1 GB	64.9 GB	4-bit small — 4-bit small quantization	Does not fit: needs 61.9 GB
61.0 GB	63 GB	69.1 GB	60.6 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Does not fit: needs 61.9 GB
65.1 GB	67 GB	73.2 GB	70.6 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Does not fit: needs 66 GB
66.4 GB	68.3 GB	74.5 GB	71.8 GB	4-bit large — 4-bit large quantization	Does not fit: needs 67.3 GB
74.5 GB	76.4 GB	82.6 GB	75.9 GB	5-bit small — 5-bit small quantization	Does not fit: needs 75.4 GB
77.2 GB	79.1 GB	85.3 GB	80.8 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Does not fit: needs 78.1 GB
78.5 GB	80.5 GB	86.6 GB	81.6 GB	5-bit large — 5-bit large quantization	Does not fit: needs 79.4 GB
89.3 GB	91.2 GB	97.4 GB	95.9 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 90.2 GB
108.1 GB	110.1 GB	116.2 GB	113.6 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 109 GB
215.8 GB	217.8 GB	224 GB	213.8 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 216.8 GB

Selected Q4_K_M

4-bit medium quantization

66.0GBat 16K context

Weights: 64.6 GB
KV cache: 1.0 GB
Overhead: 0.3 GB

Doesn't fit RTX 4090 (reference) · needs 42.0 GB more

≥30 tok/s needs ≥430 GB/s · M3 Ultra ~41, M2 Ultra ~41

Or .

Which Hardware Can Run GLM 4.6V?

Q4_K_M · 65.1 GB

GLM 4.6V (Q4_K_M) requires 65.1 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 85+ GB is recommended. Using the full 131K context window can add up to 8.1 GB, bringing total usage to 73.2 GB. No consumer GPU has enough memory. 30 devices with unified memory can run GLM 4.6V, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 31 GB8 GB tier →12 GB tier →16 GB tier →

24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 31 GB24 GB tier →

Mac 32 GB · Mac Mini M4Doesn't fit — smallest quant needs 31 GBMacs →

Mac 64 GB · Mac Studio M4 MaxIQ4_XS · 59.2 GB · ~34 tok/s · 32K59.2~3432KMacs →

How good is GLM 4.6V?

llmrun Score blends public benchmark boards into one number; 100 = the average open model.

Not enough data yet for an llmrun Score.

A Score needs results on enough independent boards. How the Composite Score is built

Benchmarks

LMArena Text
LMArena · Oct 2026Arena
1378.4
#63/188 open
Arena

GLM 4.6V benchmark results

Alternatives to GLM 4.6V

GLM 4 family · 18 models →

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Llama 4 Scout 17B 16E InstructMeta	82.1	77 GB	—needs > 24 GB	same size
C4ai Command A 03 2025Cohere	79.8	78.7 GB	—needs > 24 GB	same size
Ring Flash 2.0Inclusion AI	—	63.1 GB	—needs > 24 GB	same size
Ling Flash 2.0Inclusion AI	—	63.1 GB	—needs > 24 GB	same size
C4ai Command R Plus 08 2024Cohere	—	73.8 GB	—needs > 24 GB	same size
Qwen1.5 110BAlibaba	—	72.4 GB	—needs > 24 GB	same size

Specifications

Publisher: Z.ai
Family: GLM 4
Parameters: 107.7B
Active parameters: 14.3B per token
Architecture: Glm4vMoeForConditionalGeneration
Context Length: 131,072 tokens
Vocabulary Size: 151,552
Release Date: 2025-12-07
License: MIT

Resources

Frequently Asked Questions

How much VRAM does GLM 4.6V need?

GLM 4.6V requires 65.1 GB of VRAM at Q4_K_M, or 215.8 GB at BF16. Full 131K context adds up to 8.1 GB (73.2 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 107.7B × 4.8 bits ÷ 8 = 64.6 GB

KV Cache + Overhead ≈ 0.5 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 8.6 GB (at full 131K context)

VRAM usage by quantization

Q4_K_M

65.1 GB

Q4_K_M + full context

73.2 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run GLM 4.6V?

Yes, at IQ2_XXS (30.1 GB) or lower. Higher quantizations like IQ2_XS (32.7 GB) exceed the NVIDIA GeForce RTX 5090's 32 GB.

What's the best quantization for GLM 4.6V?

For GLM 4.6V, Q4_K_M (65.1 GB) offers the best balance of quality and VRAM usage. Q4_K_L (66.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 30.1 GB.

VRAM requirement by quantization

IQ2_XXS

30.1 GB~53%

IQ3_XS

44.9 GB~73%

Q3_K_L

55.6 GB~86%

Q4_K_M ★

65.1 GB~89%

Q4_K_L

66.4 GB~90%

BF16

215.8 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run GLM 4.6V on a Mac?

Yes — Mac Studio M5 Max (36 GB) and 65 other Macs can run GLM 4.6V, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 65.1 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).

Can I run GLM 4.6V locally?

Yes — GLM 4.6V can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is GLM 4.6V?

At Q4_K_M, GLM 4.6V can reach ~91 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 8.6 GB active ÷ (8000 × 0.65) = 1.66 ms, plus 46 layers × 0.055 ms = 2.53 ms, so 1000 ÷ 4.19 ms = ~239 tok/s

Estimated speed at Q4_K_M (65.1 GB)

NVIDIA B200

~239 tok/s

NVIDIA B300

~239 tok/s

NVIDIA GH200 Grace Hopper Superchip

~191 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of GLM 4.6V?

At Q4_K_M, the download is about 64.63 GB. The full-precision BF16 version is 215.42 GB. The smallest option (IQ2_XXS) is 29.62 GB.

Which GPUs can run GLM 4.6V?

No single consumer GPU has enough VRAM to run GLM 4.6V at Q4_K_M (65.1 GB). Multi-GPU or professional hardware is required.

Which devices can run GLM 4.6V?

30 devices with unified memory can run GLM 4.6V at Q4_K_M (65.1 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.