Best text-to-image models for a 16GB graphics card
A leaderboard of the best text-to-image models for a 16GB graphics card (with 32GB of system RAM). The best one that runs on it right now is Ideogram 4.0, followed by Qwen-Image-2.1 and Z-Image (Base). Models are ranked on human votes and image generation tests like GenEval. Models that run on it show the download that fits and the command to run it.
126 models · scores combine 41 benchmarks · updated how it works · find a leaderboard · benchmark your machine
Scores use reviewed text to image evaluations only. Compare scores within this task. Models without task measurements are excluded.
Models for 32GB RAM · 16GB graphics card → ~16GB fast · reset
+ more
$ terminal
↗ share
detecting your OS…
Score vs API price
x
show
● open weights ◆ API only ─┘ best score for the price · x: USD per image (log)
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|
| # | model | task score↓ | where↓ | speed↓ | sources |
|---|---|---|---|---|---|
| 1 |  GPT Image 2.5 Sunburst | 90±2* | api $0.013/image |   | |
| 2 | GPT Image 2.5 Flare | 88±2* | api $0.013/image | | |
| 3 | GPT Image 2 | 85±2 | api $0.053/image |  | |
| 5 |  Nano Banana 2.1 | 80±2* | api $0.034/image | | |
| 6 |  Reve 2.1 | 78±2* | api $0.25/image | | |
| 7 |  Grok Imagine Image 2.0 | 78±2* | api $0.040/image | | |
| 8 | Nano Banana 2 (Gemini 3.1 Flash Image) | 75±2 | api $0.067/image | | |
| 9 |  Muse Image | 75±2* | api $0.010/image | | |
| 10 |  Qwen-Image-3.0-Pro | 74±2 | api $0.040/image | | |
| 11 |  Seedream 5.0 Pro | 74±2 | api $0.050/image | | |
| 12 | Reve 2.0 | 74±2* | api $0.25/image | | |
| 13 | Nano Banana Pro (Gemini 3 Pro Image) | 73±2 | api $0.13/image | | |
| 15 | Nano Banana 2 Lite (Gemini 3.1 Flash-Lite Image) | 72±2* | api $0.034/image | | |
| 16 | GPT Image 1.5 | 70±2 | api $0.034/image | | |
| 17 | Qwen-Image 2.0 Pro | 69±4 | api $0.075/image | | |
| 18 |  Ideogram 4.0 · nf4 | 68±2* | local | | |
| 19 | Qwen-Image-2.1 · gguf-q4_k + Qwen3-VL 8B q4_k_m (sd.cpp) | 67±2* | local | | |
| 21 | Qwen-Image-2.0-Pro (2026-06-22) | 66±2* | api $0.075/image | | |
| 22 |  Uni-1.1 Max | 66±2* | api $0.10/image | | |
| 23 | Uni-1.1 | 66±2* | api $0.040/image | | |
| 25 | Grok Imagine Image | 64±2* | api $0.020/image | | |
| 26 |  Recraft V4.1 Utility Pro | 64±2* | api $0.21/image | | |
| 27 |  FLUX.2 [max] | 63±2 | api $0.070/image | | |
| 28 | Grok Imagine Image Pro | 63±2* | api $0.050/image | | |
| 29 | Seedream 5.0 Lite | 63±2 | api $0.035/image | | |
| 30 | FLUX.2 [flex] | 63±2* | api $0.050/image | | |
| 32 | FLUX.2 [pro] | 62±2 | api $0.030/image | | |
| 33 |  Seedream 4.5 | 62±2 | api $0.040/image | | |
| 34 |  ERNIE-Image | 62±3 | api $0.030/image +local |  | |
| 35 | Nano Banana (Gemini 2.5 Flash Image) | 62±2 | api $0.039/image |  | |
| 37 | ERNIE-Image-Turbo | 61±3† | api $0.010/image +local | | |
| 38 | GPT Image 1 | 61±1 | api $0.042/image | +3 | |
| 39 | Imagen 4 Ultra | 61±2 | api $0.060/image | | |
| 40 |  Wan 2.6 Text-to-Image | 61±2* | api $0.030/image | | |
| 41 | Seedream 4.0 | 60±1 | api $0.040/image | | |
| 42 | Recraft V4.1 Pro | 60±2* | api $0.21/image | | |
| 45 |  Krea 2 Medium | 59±2* | api $0.030/image | | |
| 46 | Wan 2.7 Image Pro | 59±2* | api $0.075/image | | |
| 47 | Imagen 4 | 59±2 | api $0.040/image | | |
| 48 | FLUX.2 [dev] | 59±1 | api $0.046/image +local | +1 | |
| 49 | Wan 2.5 Text-to-Image Preview | 58±2* | api $0.030/image | | |
| 50 | Recraft V4 | 58±2* | api $0.040/image | | |
| 51 | GPT Image 1 Mini | 58±2* | api $0.011/image | | |
| 52 | Krea 2 Turbo | 58±2* | api $0.015/image +local | | |
| 53 |  Z-Image (Base) · gguf-q4_k_m + Qwen3 4B q4_k_m (sd.cpp) | 58±2 | local | | |
| 54 | Krea 2 Large | 58±2* | api $0.060/image | | |
| 55 |  HiDream-O1-Image · fp8 | 58±2 | local | | |
| 56 |  HunyuanImage 3.0 | 58±2 | api $0.080/image | +2 | |
| 57 | Qwen-Image-2512 | 57±1 | api $0.020/image +local | | |
| 58 | Wan 2.7 Image | 57±2* | api $0.030/image | | |
| 59 | Recraft V4.1 Flash | 57±2* | api $0.0070/image | | |
| 62 | Z-Image-Turbo · bf16 | 56±1 | local | +1 | |
| 63 |  LLaDA-Image · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) | 55±3† | local (slow) | | |
| 65 | HunyuanImage 2.1 | 55±6* | api $0.10/image | | |
| 66 | FLUX.1 Kontext [max] | 54±2* | api $0.080/image | | |
| 68 |  LongCat-Image · gguf-q4_k_m + Qwen2.5-VL q8_0 (sd.cpp) | 54±4 | local | | |
| 69 | LLaDA-Image-Turbo · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) | 53±3† | local (slow) | | |
| 70 | FLUX.2 [klein] 9B · gguf-q4_0 + Qwen3 8B q4_k_m (sd.cpp) | 53±2 | local | | |
| 71 | Qwen-Image · nf4 (bitsandbytes 4-bit transformer + text encoder) | 53±1 | local | +5 | |
| 72 |  Kling Image 2.1 | 53±7* | api $0.014/image | | |
| 73 | Ideogram 3.0 | 52±2* | api $0.060/image | | |
| 74 | FLUX.1 Kontext [pro] | 52±2* | api $0.040/image | | |
| 76 | Photon | 51±2 | api $0.019/image | | |
| 77 |  P-Image | 50±2* | api $0.0050/image | | |
| 78 | FLUX.2 [klein] 4B · bf16 (cpu offload) | 49±2 | local | | |
| 79 |  Runway Gen-4 Image | 49±2* | api $0.050/image | | |
| 80 |  GLM-Image | 49±2 | api $0.015/image | | |
| 81 | Recraft V3 | 49±2 | api $0.040/image | | |
| 82 |  Kandinsky 3 · fp16 | 48±10* | local (slow) | | |
| 83 |  Lucid Origin | 48±2* | api $0.017/image | | |
| 86 |  Ovis-Image-7B · gguf-q4_0 + Ovis 2.5 (sd.cpp) | 48±7* | local | | |
| 87 | FLUX.1 Krea [dev] | 48±4* | api $0.025/image | | |
| 88 | HiDream-I1-Full | 47±2 | api $0.050/image | | |
| 89 | Ideogram 2.0 | 46±2 | api $0.080/image | | |
| 92 | FLUX1.1 [pro] | 45±2 | api $0.040/image | | |
| 95 | UNO · fp8 FLUX.1-dev base + UNO LoRA, CPU offload | 44±4* | local (slow) | | |
| 97 | FLUX.1 [dev] · bnb-8bit | 43±1 | local | +3 | |
| 98 |  Playground v2 (1024px aesthetic) · fp16 | 43±4* | local | | |
| 100 |  Lumina-Image 2.0 · bf16 + model CPU offload | 42±4* | local | | |
| 101 | FLUX.1 [schnell] | 42±3* | api $0.0030/image +local | | |
| 102 |  Show-o2 7B · official scripts (checkpoint as stored) | 42±3 | local (slow) | | |
| 104 | Playground v2.5 · fp16 | 41±2 | local | +1 | |
| 105 |  Stable Diffusion 3.5 Large · fp8 (TensorRT) | 41±1 | local | +1 | |
| 106 | BAGEL · nf4 (app.py --mode 2, bitsandbytes) | 41±1 | local | +2 | |
| 107 | FLUX.1 Kontext [dev] | 40±2* | api $0.025/image +local | | |
| 108 | Stable Diffusion 3.5 Medium | 40±3* | api $0.035/image +local | | |
| 109 | Show-o2 1.5B-HQ · official scripts (checkpoint as stored) | 40±4* | local | | |
| 111 |  OmniGen2 | 40±4* | api $0.15/image +local | | |
| 112 | Stable Cascade · bf16 prior + fp16 decoder, model CPU offload | 40±4* | local | | |
| 113 | SDXL-Lightning (4-step UNet) · fp16 | 40±4* | local | | |
| 114 | Stable Diffusion 3 Medium · q8_0 (sd.cpp --type) | 39±2 | local | | |
| 117 |  Kolors · fp16 | 38±3 | local (slow) | | |
| 119 |  PixArt-α (PixArt-XL-2-1024-MS) · fp16 | 38±3 | local | | |
| 120 |  Janus-Pro-7B | 37±2 | api $0.0020/image +local | | |
| 121 | Kandinsky 2.2 · fp16 | 37±5*† | local | | |
| 122 |  SSD-1B · fp16 (sd.cpp) | 36±3* | local | | |
| 123 | Show-o (show-o-demo-512) · fp32 | 36±3* | local | | |
| 124 | Janus-Pro-1B · bf16 | 36±4 | local | | |
| 126 | Emu3 · bf16 | 35±2 | local (slow) | | |
| 127 | Stable Diffusion XL 0.9 (base) · fp16 | 35±3* | local | | |
| 128 | SDXL-Turbo · fp16 | 35±2 | local | | |
| 129 | Stable Diffusion XL 1.0 · fp16 | 35±1 | local | +5 | |
| 130 |  IF-XL · fp16 + model CPU offload (stages I+II, 256px) | 34±2 | local (slow) | | |
| 131 | IF-L · fp16 + model CPU offload (stages I+II, 256px) | 34±4* | local | | |
| 132 |  LCM-SDXL · fp16 | 34±3* | local | | |
| 133 |  Dreamlike Photoreal 2.0 · fp16 | 34±4* | local | | |
| 134 | IF-M · fp16 + model CPU offload (stages I+II, 256px) | 33±4* | local | | |
| 135 |  Versatile Diffusion v1.0 · fp16 | 33±4* | local | | |
| 136 |  CogView2 · fp16 (--fp16), first stage only | 32±4* | local | | |
| 137 | LCM-SSD-1B · fp16 | 32±3* | local | | |
| 140 |  VILA-U 7B (256px) · bf16 (official inference.py) | 32±6* | local (slow) | | |
| 141 |  Latent Diffusion text2img-large (LDM-KL-8-G, LAION-400M) · fp32 | 32±4* | local | | |
| 143 | Janus-1.3B · bf16 | 31±4 | local | | |
| 144 |  VQ-Diffusion (ITHQ) · fp16 | 31±4* | local | | |
| 146 |  DALL·E Mega (minDALL-E) · fp32 | 31±3* | local | | |
| 149 |  Openjourney · fp16 | 29±2 | local | | |
| 150 | Stable Diffusion v1.4 · fp16 | 28±2 | local | | |
| 151 | Stable Diffusion v1.3 · fp16 | 28±4* | local | | |
| 152 | Stable Diffusion v1.1 · fp16 | 27±3 | local | | |
| 153 | Stable Diffusion v1.5 · fp16 | 27±2 | local | | |
| 154 | Stable Diffusion v1.2 · fp16 | 27±3* | local | | |
| 155 | LCM Dreamshaper v7 · fp32 | 24±4* | local | | |
| — | Kandinsky 2.1 · fp16 | not scored yet | local | ||
| — | Mage-Flow · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) | not yet calibrated | local | | |
| — | Mage-Flow-Base · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) | not yet calibrated | local | | |
| — | Mage-Flow-Turbo · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) | not yet calibrated | local | |
+ 36 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