whatmodel.sh · find the best image models you can run

Best text-to-image models to run locally on a 32GB graphics card

A leaderboard of the best text-to-image models you can run locally on a 32GB graphics card (with 32GB of system RAM). Ideogram 4.0 is #1 right now, followed by Qwen-Image-2.1 and ERNIE-Image. Models are ranked on human votes and image generation tests like GenEval. Each model shows the download that fits and the command to run it.

73 models · scores combine 41 benchmarks · updated 2026-10-11 02:20 UTChow it works · find a leaderboard · benchmark your machine

task[all tasks][text to image][image to image / editing]

Scores use reviewed text to image evaluations only. Compare scores within this task. Models without task measurements are excluded.

Models for 32GB RAM · 32GB graphics card → ~32GB fast · local only · reset

+ more

card[RTX 5090][RTX 5000 Ada][RTX PRO 4500 Blackwell]

set[all sets][LMArena text-to-image · 157][CVTG-2K word accuracy · 3]

Filter the model list while keeping task scores. Models without a shared score scale are not yet calibrated; select a set to see its own scores.

Detect your exact chip or graphics card:

curl -G whatmodel.sh/image -d max=0 --data-urlencode "mac=$(sysctl -n hw.memsize machdep.cpu.brand_string)"Mac

curl -G whatmodel.sh/image -d max=0 --data-urlencode "ram=$(awk '/MemT/{print $2}' /proc/meminfo)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>/dev/null)"Linux

curl.exe -G whatmodel.sh/image -d max=0 --data-urlencode "ram=$((gcim Win32_ComputerSystem).TotalPhysicalMemory)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>$null)"Windows

$ terminal

ram[8][16][24][32][48][64][96][128][192][256]GB

gpu[apple silicon][graphics card][none]

vram[1][2][4][8][12][16][24][32][48][64][80][96][128][160]GB

price[local][$0.001][$0.005][$0.02][$0.1][any]max per image

detecting your OS…

Score vs memory to run

     │                                    ● Ideogram 4.0       ∘
  60 ┤ · · · · · · ·HiDream-O1-Image ●────┘· · · · ∘ · · ·∘· · · ·∘·∘·∘· ·∘· · ∘ ·∘· · · · · · · · · · · ·∘· ·
     │                               ∘         ∘   ∘   ∘       ∘  ∘     ∘          ∘     ∘   ∘     ∘ ∘
  40 ┤ · ●───────────●─SSD-1B─∘──────∘ · ·∘· · ∘ · ∘ · ∘ ··· ∘ ∘ ·∘·∘· · ·∘·∘· ∘∘·∘· · · ·∘· · · · · ∘ · · · ·
     │   ∘
     └───┬───────────────────────────┬────────────────────┬────────────────────┬───────────────────────────┬──
        2GB                         5GB                 10GB                 20GB                        50GB

#	model	task score↓	where↓	speed↓	RAM↓	get it	sources

#	model	task score↓	where↓	RAM↓	get it	sources
18	Ideogram 4.0 · fp8 + Qwen3-VL 8B q4_k_m (sd.cpp) Ideogram	68±2*	local	~29GB	how to run ›
19	Qwen-Image-2.1 · int8 convrot + Qwen3-VL 8B q4_k_m (sd.cpp) Alibaba	67±2*	local	~16GB	how to run ›
34	ERNIE-Image · bf16 Baidu	62±3	local	~20GB	how to run ›
37	ERNIE-Image-Turbo · bf16 + Ministral 3B (sd.cpp) Baidu	61±3†	local	~29GB	how to run ›
48	FLUX.2 [dev] · nf4 (bnb-4bit DiT + text encoder) Black Forest Labs	59±1	local	~20GB	how to run ›	+1
52	Krea 2 Turbo · gguf-q8_0 + Qwen3-VL 4B q4_k_m (sd.cpp) Krea	58±2*	local	~20GB	how to run ›
53	Z-Image (Base) · bf16 + Qwen3 4B bf16 (sd.cpp) Alibaba (Tongyi-MAI)	58±2	local	~25GB	how to run ›
55	HiDream-O1-Image · bf16 HiDream	58±2	local	~20GB	how to run ›
56	HunyuanImage 3.0 · nf4 (community, EricRollei/HunyuanImage-3-NF4-v2) Tencent	58±2	local (slow)	~49GB	how to run ›	+2
57	Qwen-Image-2512 · gguf-q4_k_m (Unsloth) + Qwen2.5-VL Q4_K_XL Alibaba	57±1	local	~22GB	how to run ›
62	Z-Image-Turbo · bf16 Alibaba (Tongyi-MAI)	56±1	local	~15GB	how to run ›	+1
63	LLaDA-Image · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) Ant Group	55±3†	local	~22GB	how to run ›
68	LongCat-Image · gguf-q4_k_m + Qwen2.5-VL q8_0 (sd.cpp) Meituan	54±4	local	~15GB	how to run ›
69	LLaDA-Image-Turbo · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) Ant Group	53±3†	local	~22GB	how to run ›
70	FLUX.2 [klein] 9B · bf16 (cpu offload) Black Forest Labs	53±2	local	~22GB	how to run ›
71	Qwen-Image · gguf-q4_k_s Alibaba	53±1	local	~25GB	how to run ›	+5
78	FLUX.2 [klein] 4B · bf16 + Qwen3 4B bf16 (sd.cpp) Black Forest Labs	49±2	local	~20GB	how to run ›
80	GLM-Image · bf16 Z.ai	49±2	local (slow)	~39GB	how to run ›
82	Kandinsky 3 · fp16 AI Forever	48±10*	local (slow)	~32GB	how to run ›
86	Ovis-Image-7B · bf16 Alibaba (AIDC-AI)	48±7*	local	~20GB	how to run ›
87	FLUX.1 Krea [dev] · bf16 + model CPU offload Black Forest Labs / Krea	48±4*	local (slow)	~41GB	how to run ›
95	UNO · fp8 FLUX.1-dev base + UNO LoRA, CPU offload ByteDance	44±4*	local	~16GB	how to run ›
97	FLUX.1 [dev] · bf16 Black Forest Labs	43±1	local	~29GB	how to run ›	+3
98	Playground v2 (1024px aesthetic) · fp16 Playground	43±4*	local	~8.9GB	how to run ›
100	Lumina-Image 2.0 · bf16 + model CPU offload Alpha-VLLM (Shanghai AI Lab)	42±4*	local	~13GB	how to run ›
101	FLUX.1 [schnell] · bf16 Black Forest Labs	42±3*	local	~29GB	how to run ›
102	Show-o2 7B · official scripts (checkpoint as stored) Show Lab, NUS	42±3	local	~23GB	how to run ›
104	Playground v2.5 · fp16 Playground	41±2	local	~8.8GB	how to run ›	+1
105	Stable Diffusion 3.5 Large · bf16 Stability AI	41±1	local	~20GB	how to run ›	+1
106	BAGEL · nf4 (app.py --mode 2, bitsandbytes) ByteDance Seed	41±1	local	~12GB	how to run ›	+2
107	FLUX.1 Kontext [dev] · gguf-q8_0 (sd.cpp) Black Forest Labs	40±2*	local	~28GB	how to run ›
108	Stable Diffusion 3.5 Medium · bf16 Stability AI	40±3*	local	~20GB	how to run ›
109	Show-o2 1.5B-HQ · official scripts (checkpoint as stored) Show Lab, NUS	40±4*	local	~7.9GB	how to run ›
111	OmniGen2 · bf16 (official repo) BAAI	40±4*	local	~17GB	how to run ›
112	Stable Cascade · bf16 prior + fp16 decoder, model CPU offload Stability AI	40±4*	local	~13GB	how to run ›
113	SDXL-Lightning (4-step UNet) · fp16 ByteDance	40±4*	local	~8.9GB	how to run ›
114	Stable Diffusion 3 Medium · fp16 incl. CLIPs + T5 (sd.cpp) Stability AI	39±2	local	~19GB	how to run ›
116	BLIP3-o 8B · bf16 (official inference.py) Salesforce Research	39±3	local (slow)	~41GB	how to run ›
117	Kolors · fp16 Kuaishou (Kolors team)	38±3	local	~22GB	how to run ›
119	PixArt-α (PixArt-XL-2-1024-MS) · fp16 PixArt (Huawei Noah's Ark Lab et al.)	38±3	local	~14GB	how to run ›
120	Janus-Pro-7B · bf16 DeepSeek	37±2	local	~18GB	how to run ›
121	Kandinsky 2.2 · fp16 AI Forever	37±5*†	local	~12GB	how to run ›
122	SSD-1B · fp16 (sd.cpp) Segmind	36±3*	local	~5.9GB	how to run ›
123	Show-o (show-o-demo-512) · fp32 Show Lab, NUS	36±3*	local	~7.9GB	how to run ›
124	Janus-Pro-1B · bf16 DeepSeek	36±4	local	~5.5GB	how to run ›
126	Emu3 · bf16 BAAI	35±2	local	~22GB	how to run ›
127	Stable Diffusion XL 0.9 (base) · fp16 Stability AI	35±3*	local	~8.9GB	how to run ›
128	SDXL-Turbo · fp16 Stability AI	35±2	local	~8.9GB	how to run ›
129	Stable Diffusion XL 1.0 · fp16 Stability AI	35±1	local	~8.9GB	how to run ›	+5
130	IF-XL · fp16 + model CPU offload (stages I+II, 256px) DeepFloyd	34±2	local	~29GB	how to run ›
131	IF-L · fp16 + model CPU offload (stages I+II, 256px) DeepFloyd	34±4*	local	~14GB	how to run ›
132	LCM-SDXL · fp16 latent-consistency	34±3*	local	~8.9GB	how to run ›
133	Dreamlike Photoreal 2.0 · fp16 Dreamlike.art	34±4*	local	~2.6GB	how to run ›
134	IF-M · fp16 + model CPU offload (stages I+II, 256px) DeepFloyd	33±4*	local	~14GB	how to run ›
135	Versatile Diffusion v1.0 · fp16 SHI Labs	33±4*	local	~7.2GB	how to run ›
136	CogView2 · fp16 (--fp16), first stage only THUDM	32±4*	local	~14GB	how to run ›
137	LCM-SSD-1B · fp16 latent-consistency	32±3*	local	~3.6GB	how to run ›
140	VILA-U 7B (256px) · bf16 (official inference.py) MIT Han Lab / NVIDIA	32±6*	local	~20GB	how to run ›
141	Latent Diffusion text2img-large (LDM-KL-8-G, LAION-400M) · fp32 CompVis	32±4*	local	~7.5GB	how to run ›
143	Janus-1.3B · bf16 DeepSeek	31±4	local	~5.5GB	how to run ›
144	VQ-Diffusion (ITHQ) · fp16 Microsoft	31±4*	local	~4.1GB	how to run ›
146	DALL·E Mega (minDALL-E) · fp32 DALL·E Mini / minDALL-E	31±3*	local	~8GB	how to run ›
149	Openjourney · fp16 PromptHero	29±2	local	~2.6GB	how to run ›
150	Stable Diffusion v1.4 · fp16 CompVis	28±2	local	~2.6GB	how to run ›
151	Stable Diffusion v1.3 · fp16 CompVis	28±4*	local	~2.6GB	how to run ›
152	Stable Diffusion v1.1 · fp16 CompVis	27±3	local	~2.6GB	how to run ›
153	Stable Diffusion v1.5 · fp16 Stability AI	27±2	local	~4GB	how to run ›
154	Stable Diffusion v1.2 · fp16 CompVis	27±3*	local	~2.6GB	how to run ›
155	LCM Dreamshaper v7 · fp32 SimianLuo (LCM authors)	24±4*	local	~4.6GB	how to run ›
—	Kandinsky 2.1 · fp16 AI Forever	not scored yet	local	~12GB	how to run ›
—	Mage-Flow · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoft	not yet calibrated	local	~14GB	how to run ›
—	Mage-Flow-Base · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoft	not yet calibrated	local	~14GB	how to run ›
—	Mage-Flow-Turbo · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoft	not yet calibrated	local	~14GB	how to run ›

+ 89 more don't fit your machine or budget · show them

162 of 162 catalogue models listed here · sources and licences

* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory