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

Best video models to run locally on a 32GB Mac

A leaderboard of the best video models you can run locally on a Mac with 32GB of memory. Wan 2.1 14B is #1 right now, followed by Wan 2.2 TI2V 5B and Wan 2.1 T2V 1.3B. Models are ranked by one score that combines 55 published benchmarks. Each model shows the download that fits and the command to run it.

6 models · scores combine 55 benchmarks · updated how it works · find a leaderboard · benchmark your machine
task[all tasks][text to video][image to video][video editing]
Models for 32GB RAM · Apple Silicon → ~21GB fast · local only · reset
+ more
chip[M1 Pro][M1 Max][M2 Pro][M2 Max][M4][M5][M6 (24/32GB)]

Detect your exact chip or graphics card:

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

curl -G whatmodel.sh/video -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/video -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/video?ram=32&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][$0.02][$0.05][$0.1][$0.25][any]max per sec

detecting your OS…

Score vs memory to run

#modelscore↓where↓speed↓RAM↓get itsources
#modelscore↓where↓speed↓RAM↓get itsources
80Wan 2.1 14B · gguf-q4_k_m + umt5 q8_0 (sd.cpp) Alibaba47±2local~20GBhow to run › VideoPhy / VideoPhy-2 official repository leaderboardsSkyReels-V2: Infinite-length Film Generative ModelCosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)T2VPhysBench: A First-Principles Benchmark for Physical Consistency in Text-to-Video GenerationWan-AI/Wan2.1-T2V-14B model cardAlibaba Wan technical report (arXiv:2503.20314v2), VBench table (own models)
81Wan 2.2 TI2V 5B · fp16 + umt5 q8_0 (sd.cpp) Alibaba47±2local~21GBhow to run › Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)VABench: A Comprehensive Benchmark for Audio-Video Generation (arXiv:2512.09299v2)Target-Bench: Can Video World Models Achieve Mapless Path Planning with Semantic Targets?VideoArgus: Agentic Rubric-Grounded Unified Evaluation for Video Generation and Editing (arXiv:2608.05485v1)Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science DomainsPAI-Bench: A Comprehensive Benchmark For Physical AI (arXiv:2512.01989v1)+3
99Wan 2.1 T2V 1.3B · fp16 + umt5 q8_0 (sd.cpp) Alibaba40±4local~12GBhow to run › WorldJen: An End-to-End Multi-Dimensional Benchmark for Generative Video ModelsAlibaba Wan technical report (arXiv:2503.20314v2), VBench table (own models)Training-Free Efficient Video Generation via Dynamic Token Carving
107T2V-Turbo-v2 · fp32 checkpoints (official Gradio demo, MPS) UC Santa Barbara38±5*local~16GBhow to run › T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design (arXiv 2410.05677v3)WorldModelBench: Judging Video Generation Models As World Models
109T2V-Turbo (VC2) · fp32 checkpoints (official Gradio demo, MPS) UC Santa Barbara38±3†local~9.9GBhow to run › GenAI-Arena (TIGER-Lab) text-to-video leaderboardT2V-Turbo: Breaking the Quality Bottleneck of Video Consistency Model with Mixed Reward Feedback (arXiv:2405.18750v2)T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design (arXiv 2410.05677v3)
—AnimateDiff SD 1.5 v2 motion adapter · fp32 motion module + SD 1.5 (sd.cpp) AnimateDiff authorsno reviewed resultslocal~7.8GBhow to run ›

+ 136 more don't fit your machine or budget · show them

142 of 142 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