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

Best video models to run locally on a 4GB graphics card

A leaderboard of the best video models you can run locally on a 4GB graphics card (with 32GB of system RAM). LTX-2.5 is #1 right now, followed by HunyuanVideo 1.5 and Open-Sora Plan v1.3 (T2V). Models are ranked by one score that combines 55 published benchmarks. Each model shows the download that fits and the command to run it.

29 models · scores combine 55 benchmarks · updated 2026-10-11 02:20 UTChow it works · find a leaderboard · benchmark your machine

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

Models for 32GB RAM · 4GB graphics card → ~4GB fast · local only · reset

+ more

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

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][$0.02][$0.05][$0.1][$0.25][any]max per sec

detecting your OS…

Score vs memory to run

     │                                                                   ● LTX-2.5             ∘
     │   ·          Pyramid Flow (SD3)    Wan 2.2 TI2V 5B ●─────────∘────┘                     ∘          ∘
  40 ┤ · ●────────────────────●───────●──────∘────────────┘· · · · · · · · · · · · · · · · · · · · · · · ·∘· ·
     │   ∘                            ∘      ∘      ∘                    ∘                     ·
     │   ∘          ∘                 ∘                                      ∘
     └───┬──────────────────────────────────────────┬──────────────────────────────────────────┬──────────────
        5GB                                       10GB                                       20GB

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

#	model	score↓	where↓	RAM↓	get it	sources
53	LTX-2.5 · fp8 Lightricks	53±6*	local (slow)	~14GB	how to run ›
59	HunyuanVideo 1.5 · bf16 Tencent	52±2	local (slow)	~20GB	how to run ›
71	Open-Sora Plan v1.3 (T2V) · bf16 Peking University (PKU-YuanGroup)	49±7*	local (slow)	~24GB	how to run ›
80	Wan 2.1 14B · gguf-q4_k_m + umt5 q8_0 (sd.cpp) Alibaba	47±2	local (slow)	~20GB	how to run ›
81	Wan 2.2 TI2V 5B · fp16 + umt5 q8_0 (sd.cpp) Alibaba	47±2	local (slow)	~21GB	how to run ›	+3
85	CogVideoX-5B-I2V · bf16 Zhipu AI (THUDM)	46±4	local (slow)	~13GB	how to run ›
91	CogVideoX-5B · bf16 Zhipu AI (THUDM)	44±2	local (slow)	~13GB	how to run ›
96	Pyramid Flow (SD3) · bf16 Peking University / Kuaishou	42±3*	local (slow)	~8GB	how to run ›
99	Wan 2.1 T2V 1.3B · fp16 + umt5 q8_0 (sd.cpp) Alibaba	40±4	local (slow)	~12GB	how to run ›
100	Mochi 1 · bf16 Genmo	40±2	local (slow)	~25GB	how to run ›
102	Open-Sora Plan v1.3 (I2V) · bf16 Peking University (PKU-YuanGroup)	39±7*	local (slow)	~24GB	how to run ›
107	T2V-Turbo-v2 · fp32 UC Santa Barbara	38±5*	local (slow)	~9.4GB	how to run ›
109	T2V-Turbo (VC2) · fp32 UC Santa Barbara	38±3†	local (slow)	~9.4GB	how to run ›
114	CogVideoX-2B · fp16 Zhipu AI (THUDM)	36±2	local (slow)	~5.3GB	how to run ›
116	Allegro · bf16 Rhymes AI	35±2	local (slow)	~9.3GB	how to run ›
118	VideoCrafter2 · fp32 Tencent AI Lab	34±2	local (slow)	~9.4GB	how to run ›
121	Open-Sora 1.1 · bf16 HPC-AI Tech	32±7*	local (slow)	~14GB	how to run ›
122	LaVie (base T2V) · fp16 Shanghai AI Lab (Vchitect)	32±2	local (slow)	~4.9GB	how to run ›
123	LTX-Video 0.9 (2B) · bf16 Lightricks	32±3*	local (slow)	~5.3GB	how to run ›
125	NOVA-0.6B T2V · fp16 BAAI	29±3*	local (slow)	~9.3GB	how to run ›
126	ZeroScope v2 576w · fp16 cerspense	27±3	local (slow)	~7.9GB	how to run ›
127	Hotshot-XL · fp16 Hotshot	27±2	local (slow)	~10GB	how to run ›
133	VideoCrafter1 T2V · official script (checkpoint as stored) Tencent AI Lab / HKUST	26±2	local (slow)	~8.1GB	how to run ›
135	LVDM (text-to-video) · fp32 HKUST / Tencent AI Lab	25±3*	local (slow)	~6.1GB	how to run ›
136	ModelScope T2V 1.7B · fp16 Alibaba DAMO Academy	25±2	local (slow)	~4.6GB	how to run ›
138	ZeroScope v2 XL (576w → XL pipeline) · fp16 + model CPU offload (576w base, then XL vid2vid) cerspense	24±2	local (slow)	~15GB	how to run ›
—	AnimateDiff SD 1.5 v2 motion adapter · fp32 motion module + SD 1.5 (sd.cpp) AnimateDiff authors	no reviewed results	local (slow)	~7.8GB	how to run ›
—	Stable Video Diffusion (SVD, 14 frames) · fp16 · CPU offload Stability AI	no reviewed results	local (slow)	~8GB	how to run ›
—	Stable Video Diffusion XT (25 frames) · fp16 · CPU offload Stability AI	no reviewed results	local (slow)	~8GB	how to run ›

+ 113 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