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

Best image models to run locally on a 48GB Mac

A leaderboard of the best image models you can run locally on a Mac with 48GB of memory. Qwen-Image-2.1 is #1 right now, followed by Ideogram 4.0 and ERNIE-Image-Turbo. Models are ranked by one score that combines 73 published benchmarks. Each model shows the download that fits and the command to run it.

41 models · scores combine 73 benchmarks · updated how it works · find a leaderboard · benchmark your machine
task[all tasks][text to image][image to image / editing]
Models for 48GB RAM · Apple Silicon → ~36GB fast · local only · reset
+ more
chip[M3 Max (16-core CPU / 40-core GPU)][M4 Pro][M4 Max (16-core CPU / 40-core GPU)][M5 Pro][M5 Max (40-core GPU)]
set[all sets][LMArena text-to-image · 197][CVTG-2K word accuracy · 3]

Filter the model list while keeping overall 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

detecting your OS…

Run curl "whatmodel.sh/image?ram=48&gpu=apple&max=0"to get this list in your terminal (add &format=json for agents).

ram[8][16][24][32][48][64][96][128][192][256]GB
gpu[apple silicon][graphics card][none]
price[local][$0.001][$0.005][$0.02][$0.1][any]max per image

detecting your OS…

Score vs memory to run

#modelscore↓where↓speed↓RAM↓get itsources
#modelscore↓where↓speed↓RAM↓get itsources
18Qwen-Image-2.1 · int8 convrot + Qwen3-VL 8B q4_k_m (sd.cpp) Alibaba69±2local~16GBhow to run › LMArena leaderboard datasetGenExam leaderboard (OpenGVLab/GenExam README, relaxed score)
23Ideogram 4.0 · fp8 + Qwen3-VL 8B q4_k_m (sd.cpp) Ideogram66±2*local~29GBhow to run › LMArena leaderboard datasetRapidata/text-to-image-human-preference-6M (Benchmark.ai Image Benchmark) published dataset-card rankings
47ERNIE-Image-Turbo · bf16 + Ministral 3B (sd.cpp) Baidu61±3†local~29GBhow to run › ERNIE-Image model repository evaluation tables
49ERNIE-Image · gguf-ud-q4_k_m + Ministral 3B (sd.cpp, all files in memory) Baidu60±2local~17GBhow to run › ERNIE-Image model repository evaluation tablesQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)Rapidata/text-to-image-human-preference-6M (Benchmark.ai Image Benchmark) published dataset-card rankings
56Krea 2 Turbo · gguf-q8_0 + Qwen3-VL 4B q4_k_m (sd.cpp) Krea59±3*local~20GBhow to run › LMArena leaderboard dataset
59Qwen-Image-Edit-2509 · gguf-q4_k_s + Qwen2.5-VL q8_0 (sd.cpp) Alibaba59±3local~26GBhow to run › Meituan LongCat — LongCat-Image technical reportGEditBench v2: A Human-Aligned Benchmark for General Image EditingHP-Edit: A Human-Preference Post-Training Framework for Image Editing (arXiv:2604.19406v1)ViGoR-Bench: How Far Are Visual Generative Models From Zero-Shot Visual Reasoners? (arXiv:2603.25823v1)BAAI — OmniGen2 technical report
61FLUX.2 [dev] · mlx-int4 (Flux2CLI, Swift) Black Forest Labs58±1local~20GBhow to run › GenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5)LMArena leaderboard datasetVerifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward ModelingQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)TIIF-Bench: How Does Your T2I Model Follow Your Instructions? (arXiv v3)+8
63Z-Image (Base) · bf16 + Qwen3 4B bf16 (sd.cpp) Alibaba (Tongyi-MAI)58±2local~25GBhow to run › Tongyi-MAI (Alibaba) — Z-Image technical reportQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)
76Qwen-Image-2512 · gguf-q4_k_m (Unsloth) + Qwen2.5-VL Q4_K_XL Alibaba56±1local~22GBhow to run › GenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5)LMArena leaderboard datasetVerifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)Can Image Models Imagine Time? ImageTime: A Novel Benchmark for Probing Visual World Modeling Through Spatiotemporal Consistency+1
78Z-Image-Turbo · gguf-q3_k (+ Qwen3-4B Q4_K_M text encoder) Alibaba (Tongyi-MAI)55±1local~7.7GBhow to run › Tongyi-MAI (Alibaba) — Z-Image technical reportLMArena leaderboard datasetQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)Can Image Models Imagine Time? ImageTime: A Novel Benchmark for Probing Visual World Modeling Through Spatiotemporal Consistency+1
79LLaDA-Image · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) Ant Group55±3†local~22GBhow to run › LLaDA-Image: Building Strong Image Generators with Fully Open Training RecipesUltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)
80Qwen-Image-Edit · gguf-q4_k_m + Qwen2.5-VL q8_0 (sd.cpp) Alibaba55±2local~27GBhow to run › Meituan LongCat — LongCat-Image technical reportLMArena leaderboard datasetChartSync: A Benchmark for Visuo-Logical Cascading Chart Editing (arXiv:2607.10301v1)GEBench: Benchmarking Image Generation Models as GUI EnvironmentsUnicEdit-10M: A Dataset and Benchmark Breaking the Scale-Quality Barrier via Unified Verification for Reasoning-Enriched EditsHP-Edit: A Human-Preference Post-Training Framework for Image Editing (arXiv:2604.19406v1)+4
82FLUX.2 [klein] 9B · gguf-q4_0 + Qwen3 8B q4_k_m (sd.cpp) Black Forest Labs55±2local~14GBhow to run › LMArena leaderboard datasetQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)Rapidata/text-to-image-human-preference-6M (Benchmark.ai Image Benchmark) published dataset-card rankingsGEditBench v2: A Human-Aligned Benchmark for General Image EditingPaintBench: Deterministic Evaluation of Precise Visual Editing (arXiv:2606.00188v1)
85Qwen-Image · gguf-q4_k_m + Qwen2.5-VL Q4_K_M (sd.cpp) Alibaba54±1local~22GBhow to run › LMArena leaderboard datasetT2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026)ServImage: An Image Generation and Editing Benchmark from Real-world Commercial Imaging Services (arXiv:2604.24023v2)QC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)GenExam leaderboard (OpenGVLab/GenExam README, relaxed score)TIIF-Bench: How Does Your T2I Model Follow Your Instructions? (arXiv v3)+8
87LLaDA-Image-Turbo · gguf-q8_0 + LLaDA2 text encoder q4_k (sd.cpp) Ant Group54±3†local~22GBhow to run › LLaDA-Image: Building Strong Image Generators with Fully Open Training RecipesUltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)
88LongCat-Image · gguf-q4_k_m + Qwen2.5-VL q8_0 (sd.cpp) Meituan53±4local~15GBhow to run › QC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation (arXiv v4)Meituan LongCat — LongCat-Image technical report
97FLUX.2 [klein] 4B · bf16 + Qwen3 4B bf16 (sd.cpp) Black Forest Labs52±2local~20GBhow to run › LMArena leaderboard datasetQC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1)GEditBench v2: A Human-Aligned Benchmark for General Image Editing
113Ovis-Image-7B · gguf-q4_0 + Ovis 2.5 (sd.cpp) Alibaba (AIDC-AI)48±7*local~12GBhow to run › Faithful, Enriched, and Precise: Benchmarking Natural-Science Illustration Generation by T2I models (FEPBench)
126FLUX.1 Kontext [dev] · gguf-q8_0 (sd.cpp) Black Forest Labs44±1local~28GBhow to run › LMArena leaderboard datasetEdit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward ModelingMeituan LongCat — LongCat-Image technical reportHP-Edit: A Human-Preference Post-Training Framework for Image Editing (arXiv:2604.19406v1)UnicEdit-10M: A Dataset and Benchmark Breaking the Scale-Quality Barrier via Unified Verification for Reasoning-Enriched EditsGEditBench v2: A Human-Aligned Benchmark for General Image Editing+7
127FLUX.1 [dev] · bf16 Black Forest Labs44±1local~29GBhow to run › LMArena leaderboard datasetT2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026)QC-T2I-Bench: Scalable Question-Centric Text-to-Image Evaluation (Reliable Ranking, Fine-Grained Diagnosis, and Cost-Aware Routing)TIIF-Bench: How Does Your T2I Model Follow Your Instructions? (arXiv v3)Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1MMMG: A Massive, Multidisciplinary, Multi-Tier Generation Benchmark for Text-to-Image Reasoning (arXiv:2506.10963v2)+5
135FLUX.1 [schnell] · bf16 Black Forest Labs42±2local~29GBhow to run › T2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026)Rapidata/Face_Generation_Benchmark (Human Face Generation Alignment) published dataset-card leaderboardVerifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation (arXiv v4)
137Stable Diffusion 3.5 Large · bf16 + T5 fp16 (sd.cpp) Stability AI42±1local~34GBhow to run › LMArena leaderboard datasetGenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5)T2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026)MMMG: a Comprehensive and Reliable Benchmark for Multitask Multimodal Generation (arXiv:2505.17613v2)Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v1ServImage: An Image Generation and Editing Benchmark from Real-world Commercial Imaging Services (arXiv:2604.24023v2)+4
139Playground v2 (1024px aesthetic) · fp16 (sd.cpp) Playground41±4*local~8.8GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)
148Stable Diffusion XL 0.9 (base) · q8_0 (sd.cpp --type) Stability AI40±6*local~4.9GBhow to run › WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation (arXiv v4)tgxs002/HPSv2 repository README, HPS v2 benchmark table
149Stable Diffusion 3 Medium · fp16 incl. CLIPs + T5 (sd.cpp) Stability AI40±2local~19GBhow to run › T2I-ReasonBench: Benchmarking Reasoning-Informed Text-to-Image Generation (Findings of ACL 2026)EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation (arXiv v4)HPSv3 repository README, image generation benchmark table (MizzenAI/HPSv3)
156SDXL-Lightning (4-step UNet) · fp16 (sd.cpp) ByteDance37±3*local~8.8GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)SDXL-Lightning: Progressive Adversarial Diffusion Distillation (arXiv:2402.13929v3)
161SSD-1B · fp16 (sd.cpp) Segmind36±4*local~5.9GBhow to run › EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)
167Stable Diffusion XL 1.0 · fp16 (sd.cpp) Stability AI34±1local~8.8GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1EvalAlign: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models (arXiv:2406.16562v3)ImagenWorld: Stress-Testing Image Generation Models with Explainable Human Evaluation on Open-ended Real-World TasksConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (arXiv v1)+8
170SDXL-Turbo · fp16 (sd.cpp) Stability AI34±2local~8.8GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (arXiv v1)SDXL-Lightning: Progressive Adversarial Diffusion Distillation (arXiv:2402.13929v3)
178Dreamlike Photoreal 2.0 · q8_0 (sd.cpp --type) Dreamlike.art30±3local~2.1GBhow to run › EvalAlign: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models (arXiv:2406.16562v3)HEIM: Holistic Evaluation of Text-to-Image Models (arXiv 2311.04287v1)tgxs002/HPSv2 repository README, HPS v2 benchmark table
183DALL·E Mega (minDALL-E) · fp32 DALL·E Mini / minDALL-E29±4local~8GBhow to run › TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering (arXiv v3)HEIM: Holistic Evaluation of Text-to-Image Models (arXiv 2311.04287v1)tgxs002/HPSv2 repository README, HPS v2 benchmark table
185Openjourney · q8_0 (sd.cpp --type) PromptHero28±2local~2.1GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)EvalAlign: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models (arXiv:2406.16562v3)ImagenHub: Standardizing the evaluation of conditional image generation models (arXiv v4)HEIM: Holistic Evaluation of Text-to-Image Models (arXiv 2311.04287v1)tgxs002/HPSv2 repository README, HPS v2 benchmark table
186Stable Diffusion v1.4 · fp32 CompVis27±2local~8GBhow to run › GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1EvalAlign: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models (arXiv:2406.16562v3)ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (arXiv v1)ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation (arXiv 2304.05977v4)HEIM: Holistic Evaluation of Text-to-Image Models (arXiv 2311.04287v1)tgxs002/HPSv2 repository README, HPS v2 benchmark table
187Stable Diffusion v1.3 · fp32 CompVis27±4*local~8GBhow to run › GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1
188Stable Diffusion v1.1 · fp32 CompVis26±3local~8GBhow to run › GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering (arXiv v3)REAL: Realism Evaluation of Text-to-Image Generation Models for Effective Data Augmentation (arXiv:2502.10663v1)
189Stable Diffusion v1.2 · fp32 CompVis26±3*local~8GBhow to run › GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)
191Stable Diffusion v1.5 · fp32 Stability AI26±2local~8GBhow to run › GenEval authors' results: djghosh13/geneval README and arXiv 2310.11513v1EvalAlign: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models (arXiv:2406.16562v3)EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation (arXiv v2)TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering (arXiv v3)HEIM: Holistic Evaluation of Text-to-Image Models (arXiv 2311.04287v1)Evaluating Numerical Reasoning in Text-to-Image Models (GeckoNum, arXiv v3)+1
193LCM Dreamshaper v7 · openvino-fp16 (OpenVINO GenAI) SimianLuo (LCM authors)23±4*local~3GBhow to run › GenAI Arena: An Open Evaluation Platform for Generative Models (arXiv v3)
—Mage-Flow · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoftnot yet calibratedlocal~14GBhow to run › Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing
—Mage-Flow-Base · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoftnot yet calibratedlocal~14GBhow to run › Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing
—Mage-Flow-Turbo · bf16 + Qwen3-VL 4B q4_k_m (sd.cpp) Microsoftnot yet calibratedlocal~14GBhow to run › Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing

+ 161 more don't fit your machine or budget · show them

202 of 202 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