30.5BTier 100Needs ~22.8 GB
Apple
MacBook Air M4 24GB
M4LaptopM4UNIFIEDMetal
24GB
Unified Memory
120GB/s
Bandwidth
22W TDP$1,099 MSRP
Operating mode
Choose the run profile you want to optimize
Apple Silicon can fit a lot thanks to unified memory. This selector changes which serving posture we optimize for when surfacing the best local LLMs for this Mac.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
Best Local LLMs for MacBook Air M4 24GB
Apple Silicon local AI performance. Excellent for local AI. Your MacBook Air M4 24GB with 24 GB unified memory can run 76 models natively, 181 more with limits. The best match is Qwen 3.5 9B at 16 tok/s for interactive local LLM use.
76
Run great
257
Total compatible
24B
Max parameters
16
Best tok/sLOW SAMPLE
Comparison guide
Best Local LLMs for MacBook Air M4 24GB — full ranked guide
Top models ranked for coding, chat, and writing with FAQ and buyer guidance — the comparison-intent companion to this spec sheet.
Cost vs cloud API
On par with cloud API pricing — local wins on privacy + latency
Assumes 4 hours/day of active inference at 16 tok/s, MacBook Air M4 24GB amortized over 36 months, US residential electricity ($0.15/kWh), blended cloud pricing at $10 per 1M tokens (GPT-4o / Claude Sonnet tier).
6.7M
Tokens/month at this pace
$36.4
Monthly local cost
$67.4
Same tokens on cloud API
$5.39
Local $/1M tokens
Break-even: amortizes in 19.4 months vs cloud API. Price reference: $1.3k MSRP.
Quick picks
Best Local LLMs by Task
Top recommendations for common local AI workloads on your MacBook Air M4 24GB
About MacBook Air M4 24GB for AI
MacBook Air M4 24GB with 24 GB unified memory. Fourth-generation Apple Silicon with enhanced Neural Engine and improved memory bandwidth, designed for AI-first workflows including local LLM inference.
All 374 models tested
Model Compatibility Tiers
Every model ranked by how well it runs on your MacBook Air M4 24GB, grouped by fit quality
Runs Great (76 models)
These models fit comfortably and run at full speed on your Mac.
Runs with Limits (191 models)
These models run but may need quantization or have reduced context windows.
Won't Fit (107 models)
These models are too large for your Mac's unified memory.
Beyond LLMs
AI Capability Matrix
What AI tasks this Mac can handle — from text generation to image and video creation.
| Capability | Status | Representative Model | Detail |
|---|---|---|---|
| LLM Chat (7B) | Runs natively | Llama 3.1 8B Q4 | — |
| LLM Coding (30B) | Needs offload | Qwen 3 30B Q4 | — |
| LLM Large (70B) | Won’t fit | Llama 3.1 70B Q4 | — |
| Image Gen (SDXL) | Runs natively | SDXL 1.0 FP16 | ~~34.6s per image |
| Image Gen (Flux) | Won't fit | Flux.1 Dev FP16 | ~~2m 36s per image |
| Image Gen (SD 3.5) | Runs with sequential offload | SD 3.5 Large FP16 | ~~8m 34s per image |
| Video Short (25f) | Runs natively | LTX Video 2B | ~~1m 30s/frame |
| Video Long (100f) | Won't fit | Wan Video 14B | ~~1m 29s/frame |
Same chip, more memory
Upgrade to More Memory? Here's What You Gain
Compare M4 configurations to see which models become available
MacBook Pro M4 Pro 24GB
24 GB unified memory
78
Run great
257
Total fit
MacBook Pro M4 32GB
32 GB unified memory
89
Run great
292
Total fit
Unlocks: Qwen3-Coder 30B A3B Instruct, Qwen 3.5 27B
ultra-efficientportableentry-level
Specifications
Compute
ArchitectureM4
Memory
Unified Memory24 GB
Bandwidth120 GB/s
TypeUnified LPDDR5X
General
FamilyM4
SegmentLaptop
InterconnectUNIFIED
Compute PlatformMETAL
MSRP$1,099
TDP22W
Key Features
M4 chip (2nd-gen 3nm TSMC)24 GB unified memory (shared CPU/GPU/Neural Engine)120 GB/s memory bandwidth16-core Neural EngineMetal 3 GPU compute (MLX framework)Fanless design — silent inference
For AI Workloads
Strengths
- Enhanced 16-core Neural Engine for ML acceleration
- Up to 546 GB/s memory bandwidth (Max)
- Excellent power efficiency for sustained inference
- Best-in-class MLX performance
- Thunderbolt 5 for external GPU expansion
Considerations
- Maximum 128 GB unified memory (less than some workstations)
- No CUDA support — limited to MLX and llama.cpp Metal
Architecture
M4
Apple M4 is the latest Apple Silicon generation, using TSMC's second-generation 3nm process. It features an enhanced Neural Engine with up to 38 TOPS and higher memory bandwidth across all tiers.
AI Relevance
The M4 Max with 128 GB unified memory and up to 546 GB/s bandwidth is currently the fastest Apple Silicon option for local LLM inference. Combined with MLX framework optimizations, it delivers the best tokens-per-second of any Mac configuration.
Process: TSMC 3nm (2nd gen)Platform: METALPrecisions: FP32, FP16
M4 is Apple's most AI-capable chip yet with up to 546 GB/s bandwidth in the Max variant. The unified memory architecture means models up to ~90 GB (at 72% usable) can run natively without offloading, covering most 70B models at Q4 quantization.
All workloads
Recommendations by Workload
The best local LLM for each task on your MacBook Air M4 24GB
Chat
SQwen 3.5 9B
This model is a direct match for chat. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.
15.6 tok/s · 60K ctx · llama.cpp
10.1 GB / 24.0 GB Unified Memory
Coding
SQwen 3.5 9B
This model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.
15.6 tok/s · 60K ctx · llama.cpp
11.2 GB / 24.0 GB Unified Memory
Agentic Coding
SQwen 3.5 9B
This model is still usable for agentic-coding, but it is not the most specialized pick. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.
15.6 tok/s · 60K ctx · llama.cpp
13.4 GB / 24.0 GB Unified Memory
Reasoning
SQwen 3 14B
This model is a direct match for reasoning. It belongs to a current frontier family for local AI. It should run, but memory headroom will be limited. Known channels: huggingface, ollama, lm-studio.
9.6 tok/s · 34K ctx · llama.cpp
14.5 GB / 24.0 GB Unified Memory
RAG
AGranite 4.1 8B
This model is a direct match for rag. It sits in the middle of the current model mix. It fits natively with comfortable headroom. Known channels: huggingface, ollama.
17.5 tok/s · 58K ctx · llama.cpp
13.3 GB / 24.0 GB Unified Memory
Just out of reach
Models you could run with an upgrade
High-quality models that need a bit more memory
397BTier 100Needs ~247.1 GB
123BTier 100Needs ~81.2 GB
Runs on NVIDIA DGX Spark 128GB
1000BTier 100Needs ~617.2 GB
1000BTier 100Needs ~617.2 GB
Image & Video Generation
Diffusion Model Compatibility
28 of 52 models can generate images or video on your MacBook Air M4 24GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | ~4.3s | S |
| Stable Diffusion 1.5Image | 512×768 | ~8.7s | S |
| Realistic Vision v5.1Image | 512×768 | ~8.7s | S |
| DreamShaper 8Image | 512×768 | ~8.7s | S |
| LCM DreamShaper v7Image | 512×768 | ~2.6s | S |
| PixArt-SigmaImage | 1024×1024 | ~34.6s | S |
| SDXL TurboImage | 512×512 | ~4.3s | S |
| SDXL LightningImage | 1024×1024 | ~13s | S |
| Stable Diffusion XL 1.0Image | 1024×1024 | ~34.6s | S |
| Playground v2.5Image | 1024×1024 | ~51.9s | S |
| RealVisXL v5.0Image | 1024×1024 | ~39s | S |
| DreamShaper XLImage | 1024×1024 | ~39s | S |
| Juggernaut XL v9Image | 1024×1024 | ~39s | S |
| Animagine XL 3.1Image | 1024×1024 | ~39s | S |
| Pony Diffusion V6 XLImage | 1024×1024 | ~39s | S |
| Animagine XL 4.0Image | 1024×1024 | ~39s | S |
| Illustrious XLImage | 1024×1024 | ~39s | S |
| Wan Video 2.1 1.3BVideo | 256×256 | ~25.3s/frame | S |
| Stable Diffusion 3.5 MediumImage | 1024×1024 | ~1m 1s | S |
| Flux.2 Klein 4BImage | 256×256 | ~23.4s | S |
| LTX Video 2BVideo | 256×256 | ~1m 30s/frame | S |
| KolorsImage | 256×256 | ~3m 4s | S |
| Stable CascadeImage | 1024×1024 | ~1m 27s | B |
| AuraFlow v0.3Image | 1024×1024 | ~2m 36s | B |
| Stable Diffusion 3.5 LargeImage | 256×256 | ~8m 34s | B |
| Stable Diffusion 3.5 Large TurboImage | 256×256 | ~1m 34s | B |
| CogVideoX 2BVideo | 256×256 | ~1m 30s/frame | D |
| Z-Image TurboImage | 256×256 | ~1m 12s | D |
| FramePack I2VVideo | 256×256 | ~1m 4s/frame | F |
| HunyuanVideoVideo | 256×256 | ~1m 4s/frame | F |
| ChromaImage | 256×256 | ~34.6s | F |
| Flux.1 DevImage | 256×256 | ~2m 36s | F |
| Flux.1 SchnellImage | 256×256 | ~30.3s | F |
| LTX Video 13BVideo | 256×256 | ~1m 4s/frame | F |
| Flux.1 Kontext DevImage | 256×256 | ~2m 53s | F |
| AnimateDiff v1.5.3Video | 512×768 | ~15.8s/frame | F |
| Cosmos Diffusion 7BVideo | 256×256 | ~1m 36s/frame | F |
| CogVideoX 5BVideo | 256×256 | ~43.4s/frame | F |
| Wan2.2 TI2V 5BVideo | 256×256 | ~43.4s/frame | F |
| Flux.2 Klein 9BImage | 256×256 | ~17.3s | F |
| Flux.1 Fill DevImage | 256×256 | ~2m 27s | F |
| Mochi 1 PreviewVideo | 256×256 | ~57.2s/frame | F |
| HunyuanVideo 1.5Video | 256×256 | ~53.1s/frame | F |
| Helios 14BVideo | 256×256 | ~1m 6s/frame | F |
| SkyReels V2 14BVideo | 256×256 | ~1m 6s/frame | F |
| Wan Video 2.1 14BVideo | 256×256 | ~1m 6s/frame | F |
| Wan Video 2.2 14BVideo | 256×256 | ~1m 6s/frame | F |
| Qwen ImageImage | 256×256 | ~58.3s | F |
| Qwen Image EditImage | 256×256 | ~58.3s | F |
| Flux.2 DevImage | 256×256 | ~27m 18s | F |
| MAGI-1Video | 256×256 | ~1m 21s/frame | F |
| HunyuanImage 3.0Image | 256×256 | ~1m 43s | F |
Image models estimated at 1024×1024 (28 steps, FP16). Video models estimated at 768×512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Get started in 2 minutes
Run Local AI on Your MacBook Air M4 24GB
Everything you need to start running models locally with Metal acceleration and Apple Silicon unified memory
1
Install Ollama
Ollama runs natively on macOS with Metal GPU acceleration. One command to install.
curl -fsSL https://ollama.com/install.sh | sh2
Pull your first model
Qwen 3.5 9B is the best match for your MacBook Air M4 24GB. Pull and run it:
ollama run qwen3.5:9bWhat to expect: With 24 GB unified memory, your top models will run at 16-18-35 tokens/sec — fast enough for interactive chat and local LLM workflows. Cloud APIs like ChatGPT typically stream at 30-60 tok/s, so Apple Silicon is competitive for many models when the fit is good.
Upgrade paths
Upgrade from MacBook Air M4 24GB
See what you unlock with more unified memory
Upgrade options
Upgrade options
360 GB/s (+240)
B
Unlocks 12 additional models that do not fit on the current setup.Unlocks Qwen3-Coder 30B A3B Instruct, Qwen 3.5 27B, Qwen3-VL 30B A3B Instruct+9 more · +145% faster avg
Unlocks 12 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 145%.
~$1,250 MSRP
MacBook Pro M1 Pro 32GBApple upgrade
32 GB Unified (+8)200 GB/s (+80)
B
Unlocks 29 additional models that do not fit on the current setup.Unlocks Qwen3-Coder 30B A3B Instruct, Qwen 3.5 27B, Qwen3-VL 30B A3B Instruct+26 more · +32% faster avg
Unlocks 29 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 32%.
~$1,999 MSRP
456 GB/s (+336)
A
Unlocks 34 additional models that do not fit on the current setup.Unlocks Qwen3-Coder 30B A3B Instruct, Qwen 3.5 27B, Qwen3-VL 30B A3B Instruct+31 more · +111% faster avg
Unlocks 34 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 111%.
~$599 MSRP
AMD Instinct MI350X 288GBBiggest leap
288 GB VRAM (+264)8000 GB/s (+7880)
B
Unlocks 79 additional models that do not fit on the current setup.Unlocks Qwen3-Coder 30B A3B Instruct, Qwen 3.5 397B A17B, Devstral 2 123B Instruct+76 more · +758% faster avg
Unlocks 79 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 758%.
~$8,000 MSRP
Frequently Asked Questions
Compare with similar
Related guides