llmrun
Z.aiGLM 4Glm4MoeForCausalLM

GLM 4.5 Air — Hardware Requirements & GPU Compatibility

ChatFunctions

GLM-4.5-Air is Z.ai's more compact model in the GLM-4.5 series, a foundation model built for intelligent agents with 106 billion total parameters and about 12 billion active via its Mixture-of-Experts design. Like its larger sibling GLM-4.5, it offers both a thinking mode for complex reasoning and tool use and a non-thinking mode for quick replies, unifying reasoning, coding, and agentic capabilities in one checkpoint. It supports a 128K token context window and is released under the MIT license for commercial and research use. With around 106 billion total parameters, running it locally at 4-bit needs roughly 60 GB of memory, putting it in reach of a multi-GPU workstation or a high-memory Mac rather than a single card.

110.5B params · MoE15B active131K contextMIT725.3K downloads646 likes33.2K quant downloads

Doesn't fit a 24 GB GPU

Smallest consumer fit → RTX 5090 · IQ2_XXS · 31.7 GB · ~164 tok/s

Macs from36 GB IQ2_XXS · ~45 tok/s

Score 96.3#62 of 100 open#30 of 60 that fit ≤96 GBMIT

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

96.3

Benchmarks · tier chart ↓
Among 100 open modelsHow the Composite Score is built

Run GLM 4.5 Air

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Recommended

Ollama
one command · needs Ollama installed
$ollama run hf.co/unsloth/GLM-4.5-Air-GGUF:Q4_K_M

Q4_K_M needs offload on the RTX 4090 · 43.6 GB short · 16K context

Download: Q4_K_M/GLM-4.5-Air-Q4_K_M-00001-of-00002.gguf72.98 GB + 1 partunsloth

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX PRO 6000 (96 GB) (28 GB headroom) · Q4_K_M

Generation speed
~97 tok/s
generation speed
Cost per 1M output tokens
$3.95
per 1M output tokens
Deployon Vast.aiBest optionfrom $1.38/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does GLM 4.5 Air Need?

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

QuantizationVRAM+32KOn RTX 4090 (reference)
47.4 GB49.3 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsDoes not fit: needs 48.3 GB
48.8 GB50.7 GB3-bit small — 3-bit small quantizationDoes not fit: needs 49.7 GB
54.3 GB56.2 GB3-bit medium — 3-bit medium quantizationDoes not fit: needs 55.2 GB
55.7 GB57.6 GB4-bit · legacy — 4-bit legacy quantizationDoes not fit: needs 56.6 GB
66.7 GB68.6 GB4-bit medium — 4-bit medium quantization — most popular sweet spotDoes not fit: needs 67.6 GB
79.1 GB81.1 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffDoes not fit: needs 80 GB
91.6 GB93.5 GB6-bit, very good — 6-bit quantization, very good qualityDoes not fit: needs 92.5 GB
110.9 GB112.8 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 111.8 GB

Selected Q4_K_M

4-bit medium quantization

67.6GBat 16K context

Weights
66.3 GB
KV cache
1.0 GB
Overhead
0.3 GB

Doesn't fit RTX 4090 (reference) · needs 43.6 GB more

≥30 tok/s needs ≥451 GB/s · M3 Ultra ~40, M2 Ultra ~40

Or rent RTX PRO 6000 from $1.38/hr.

Which Hardware Can Run GLM 4.5 Air?

Q4_K_M · 66.7 GB

GLM 4.5 Air (Q4_K_M) requires 66.7 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 87+ GB is recommended. Using the full 131K context window can add up to 8.1 GB, bringing total usage to 74.8 GB. No consumer GPU has enough memory. 30 devices with unified memory can run GLM 4.5 Air, 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
8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 31.7 GB8 GB tier →12 GB tier →16 GB tier →
24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 31.7 GB24 GB tier →
Mac 32 GB · Mac Mini M4Doesn't fit — smallest quant needs 31.7 GBMacs →
Mac 64 GB · Mac Studio M4 MaxQ3_K_L · 57.9 GB · ~34 tok/s · 32KMacs →

How good is GLM 4.5 Air?

The llmrun Score merges 4 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the ≤96 GB tier

  • GLM 4.5 Air
  • Other models
  • Best Score for its VRAM

#30 of 60 that fit ≤96 GB

Each dot is a model at the best quant that fits ≤96 GB. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤96 GB ranking →

Benchmarks

Here are all 5 boards that have results for this model, sorted by their weight in the llmrun Score; badges show who ran the test.

  • LMArena Text

    LMArena · Oct 2026Arena

    1373.0

    #63/187 open

    Arena
  • Humanity's Last Exam

    Epoch AI · Jul 2025Lab-reported

    8.1%

    #3/4 open

    Lab-reported
  • MMLU-Pro

    TIGER-LabIndependent

    81.4%

    #19/145 open

    Independent
  • GSO-Bench

    Epoch AI · Jul 2025Lab-reported

    2.9%

    #3/3 open

    Lab-reported
  • ForecastBench

    Epoch AI · Jul 2025Lab-reported

    59.2

    #9/26 open

    Lab-reported
All 5 GLM 4.5 Air benchmarks

In the ≤96 GB class

60 models fit ≤96 GB

Same Score, only models that fit ≤96 GB at their best quant: the ranking a ≤96 GB owner actually chooses from.

  1. #1

    MiMo V2.6 Flash RLXiaomi

    131.9

    ~153 tok/s · 93.8 GB

  2. #2

    DeepSeek V4 Flash 0731DeepSeek

    127.6

    ~103 tok/s · 95.5 GB

  3. #3

    Qwen3.8 27BAlibaba

    122.1

    ~67 tok/s · 17.9 GB

  4. #4

    Inkling SmallThinking Machines

    119.6

    ~356 tok/s · 92.9 GB

  5. #5

    MiniMax M2.5MiniMax

    111.8

    ~138 tok/s · 93.9 GB

  6. #6

    Qwen3.6 27BAlibaba

    111.2

    ~67 tok/s · 17.9 GB

Full ≤96 GB ranking →RTX PRO 6000 Blackwell Workstation Edition page →

Alternatives to GLM 4.5 Air

GLM 4 family · 18 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Similar models

10
ModelScoreVRAMtok/sWhy it is similar
Qwen3.5 122B A10BAlibaba109.475.5 GB—needs > 24 GBsimilar score
Qwen3 30B A3B Instruct 2507Alibaba96.719.4 GB~173similar score
GPT OSS 20BOpenAI96.313.1 GB~190similar score
QwQ 32BAlibaba96.223 GB~34similar score
NVIDIA Nemotron 3 Super 120B A12B BF16NVIDIA95.874.7 GB—needs > 24 GBsimilar score
Llama 4 Scout 17B 16E InstructMeta81.877 GB—needs > 24 GBsimilar score
Qwen2.5 72B InstructAlibaba81.749.3 GB—needs > 24 GBsimilar score
C4ai Command A 03 2025Cohere79.878.7 GB—needs > 24 GBsame size
Llama 3.3 70B InstructMeta79.751.3 GB—needs > 24 GBsame size
Llama 3.1 70B InstructMeta77.951.3 GB—needs > 24 GBsame size
Fine-tunes & variants(2)

Specifications

Publisher
Z.ai
Family
GLM 4
Parameters
110.5B
Active parameters
15B per token
Architecture
Glm4MoeForCausalLM
Context Length
131,072 tokens
Vocabulary Size
151,552
Release Date
2025-07-20
License
MIT

Resources

Frequently Asked Questions

How much VRAM does GLM 4.5 Air need?

GLM 4.5 Air requires 66.7 GB of VRAM at Q4_K_M, or 221.4 GB at BF16. Full 131K context adds up to 8.1 GB (74.8 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 110.5B × 4.8 bits ÷ 8 = 66.3 GB

KV Cache + Overhead ≈ 0.4 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 ≈ 8.5 GB (at full 131K context)

VRAM usage by quantization

66.7 GB
74.8 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run GLM 4.5 Air?

Yes, at IQ2_XXS (30.8 GB) or lower. Higher quantizations like IQ2_XS (33.6 GB) exceed the NVIDIA GeForce RTX 5090's 32 GB.

What's the best quantization for GLM 4.5 Air?

For GLM 4.5 Air, Q4_K_M (66.7 GB) offers the best balance of quality and VRAM usage. Q5_K_S (76.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 30.8 GB.

VRAM requirement by quantization

IQ2_XXS
30.8 GB
IQ3_XS
46.0 GB
Q3_K_L
57.0 GB
Q4_K_M ★
66.7 GB
Q5_K_S
76.4 GB
BF16
221.4 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run GLM 4.5 Air on a Mac?

Yes — Mac Studio M5 Max (36 GB) and 65 other Macs can run GLM 4.5 Air, 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 66.7 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).

Can I run GLM 4.5 Air locally?

Yes — GLM 4.5 Air can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is GLM 4.5 Air?

At Q4_K_M, GLM 4.5 Air can reach ~90 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 → 9.0 GB active ÷ (8000 × 0.65) = 1.74 ms, plus 46 layers × 0.055 ms = 2.53 ms, so 1000 ÷ 4.27 ms = ~234 tok/s

Estimated speed at Q4_K_M (66.7 GB)

~234 tok/s
~234 tok/s
~186 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 4.5 Air?

At Q4_K_M, the download is about 66.28 GB. The full-precision BF16 version is 220.94 GB. The smallest option (IQ2_XXS) is 30.38 GB.

Which GPUs can run GLM 4.5 Air?

No single consumer GPU has enough VRAM to run GLM 4.5 Air at Q4_K_M (66.7 GB). Multi-GPU or professional hardware is required.

Which devices can run GLM 4.5 Air?

30 devices with unified memory can run GLM 4.5 Air at Q4_K_M (66.7 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.