Best image models for a 24 GB Mac
Which open-weight image models run on a 24 GB Mac. Verdicts come from exact Hugging Face file sizes, model card and app maker statements, and published runs.
- Run
- 5
- Unknown
- 4
- Does not fit
- 1
5 of 10 listed image models run on a 24 GB Mac: Qwen-Image-2.1, FLUX.1 [dev], Stable Diffusion 3.5 Large, FLUX.2 [klein] and SDXL Base 1.0. Highest blind-vote rank among them: Qwen-Image-2.1, #19 of 82 in text to image (Arena, snapshot 2026-10-07). No figure covers FLUX 3 Image Klein, Krea 2 Turbo, MOSS-Image and Z-Image-Turbo at this size yet. FLUX.2 [dev] needs more memory.
License check before commercial use: Qwen-Image-2.1, FLUX.1 [dev], Stable Diffusion 3.5 Large, FLUX.2 [klein].
Get the weekly local AI newsletter
The Thursday ModelFit newsletter: new local models and what fits your machine.
For: MacBook Air M5 16 GB
You'll also get the weekly ModelFit email. By signing up you agree to our Privacy Policy. Unsubscribe from either email anytime, in one click.
Which image models run on a 24 GB Mac?
10 modelsRuns means a published run on 24 GB or less, or a model card or app maker figure at or below it. Does not fit means the weights the app keeps resident in the smallest complete build are larger than 24 GB. A Mac has one memory pool, so the text encoder cannot move out to a separate system RAM; an app that loads the text encoder and the main model one after the other counts only its largest stage. unknown means no source covers this size yet; it is not a failure.
| Model | Verdict | Why | Quality | App |
|---|---|---|---|---|
| Qwen-Image-2.1Qwen Research License Agreement | Runs · Stated | Stated by the source: 16 GB of unified memory runs it, below this 24 GB tier (mlx-serve 4-bit pack, text encoder freed before the denoise).mlx-serve changelog | #19 of 82 in text to image11,441 votes | Diffusers with staged loading, Diffusers, stable-diffusion.cpp |
| FLUX.1 [dev]FLUX.1 [dev] Non-Commercial License v1.1.1 | Runs · Published | A dated run on Apple M3 Pro with 18 GB of unified memory, below this 24 GB tier: Draw Things 5-bit, 1024x1024.Draw Things engineering: Metal FlashAttention 2.0 | #78 of 82 in text to image49,221 votes | Diffusers, mflux, ComfyUI |
| Stable Diffusion 3.5 LargeStability AI Community License | Runs · Published | A dated run on Apple M3 Pro with 18 GB of unified memory, below this 24 GB tier: Draw Things 5-bit, 1024x1024.Draw Things engineering: Metal FlashAttention 2.0 | #81 of 82 in text to image23,396 votes | ComfyUI, Rapid-MLX |
| FLUX.2 [klein]Apache 2.0 (4B), FLUX Non-Commercial License (9B) | Runs · Stated | Stated by the source: 24 GB of unified memory runs it (flux-2-swift-mlx, Klein 9B qint8).flux-2-swift-mlx README | No independent score yet | flux-2-swift-mlx, Diffusers, stable-diffusion.cpp |
| SDXL Base 1.0CreativeML Open RAIL++-M License | Runs · Published | A dated run on Apple M2 Pro with 16 GB of unified memory, below this 24 GB tier: ComfyUI fp16, 1344x768, 20 steps.ComfyUI issue #948 | No independent score yet | ComfyUI, Rapid-MLX, stable-diffusion.cpp |
| FLUX 3 Image KleinFLUX Kommunity License v1.0 | unknown | The smallest complete build, mflux 4-bit, is a 20.3 GB download, but no stated figure or published run covers 24 GB of unified memory yet. | No independent score yet | Diffusers, mflux |
| Krea 2 TurboKrea 2 Community License Agreement v.1 | unknown | The smallest complete build, Turbo MLX mixed-4/8, is a 19.2 GB download, but no stated figure or published run covers 24 GB of unified memory yet. | No independent score yet | ComfyUI-Krea2-MLX, Diffusers, stable-diffusion.cpp |
| MOSS-ImageApache 2.0 | unknown | The smallest complete build, Original SFT safetensors, is a 27.2 GB download, but no stated figure or published run covers 24 GB of unified memory yet. | No independent score yet | MOSS-Image MPS path, stable-diffusion.cpp |
| Z-Image-TurboApache 2.0 | unknown | The smallest complete build, Q4_K_M GGUF, is a 12.1 GB download, but no stated figure or published run covers 24 GB of unified memory yet. | #59 of 82 in text to image24,808 votes | Diffusers, mflux, stable-diffusion.cpp |
| FLUX.2 [dev]FLUX Non-Commercial License | Does not fit est. | The smallest complete build, Q4_K_M GGUF, keeps 34.8 GB of weights resident, more than 24 GB of unified memory. | #38 of 82 in text to image77,988 votes | flux-2-swift-mlx, Diffusers, ComfyUI, stable-diffusion.cpp |
Quality: blind human votes from Arena (arena.ai), CC BY 4.0, snapshot 2026-10-07. Scores are for the Arena checkpoint, not your local quantized build.
Frequently asked questions
What is the best image model for a 24 GB Mac?
Qwen-Image-2.1 has the highest blind-vote rank among the image models that run on a 24 GB Mac: #19 of 82 in text to image, from 11,441 votes (Arena, snapshot 2026-10-07). FLUX.1 [dev], Stable Diffusion 3.5 Large, FLUX.2 [klein] and SDXL Base 1.0 also run.
Which image models run on a 24 GB Mac?
Qwen-Image-2.1 (16 GB stated by mlx-serve changelog), FLUX.1 [dev] (run on Apple M3 Pro with 18 GB), Stable Diffusion 3.5 Large (run on Apple M3 Pro with 18 GB), FLUX.2 [klein] (24 GB stated by flux-2-swift-mlx README) and SDXL Base 1.0 (run on Apple M2 Pro with 16 GB).
Which image models need more than 24 GB on a Mac?
FLUX.2 [dev] (smallest complete build Q4_K_M GGUF, 34.8 GB). Its resident weights alone are larger than the memory.