AI models that fit on a 8 GB GPU
Open models that fit in 8 GB of VRAM: 540 as published, 704 at FP8 and 1048 at INT4. Each model is listed on the smallest tier it fits at that precision.
GPUs with 8 GB
Cards whose datasheet VRAM puts them in this tier, up to the next tier at 12 GB. Prices are the lowest live data-center on-demand rate per GPU.
| GPU | VRAM | From per GPU hour |
|---|---|---|
| RTX 3060 Ti | 8 GB | No live offer |
| RTX 3070 | 8 GB | No live offer |
| RTX 3070 Ti | 8 GB | No live offer |
| RTX 4060 | 8 GB | No live offer |
| RTX 5060 | 8 GB | No live offer |
| RTX 3080 | 10 GB | No live offer |
| GTX 1080 Ti | 11 GB | No live offer |
| RTX 2080 Ti | 11 GB | No live offer |
Fits as published
Models whose published weights, plus the flat overhead, fit this much memory with no quantization.
Text models
The 40 most downloaded of 244.
Vision-language models
The 40 most downloaded of 60.
Image generation models
The 40 most downloaded of 47.
Video generation models
10 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| Wan2.1-T2V-1.3B-Diffusers | 1.4B | F32 | 6.3 GB |
| stable-video-diffusion-img2vid-xt | 1.5B | F32 | 6.8 GB |
| TurboWan2.1-T2V-1.3B-Diffusers | 1.4B | BF16 | 3.2 GB |
| stable-video-diffusion-img2vid | 1.5B | F32 | 6.8 GB |
| i2vgen-xl | 1.4B | F32 | 6.3 GB |
| Wan2.1-T2V-1.3B | 1.4B | F32 | 6.3 GB |
| CogVideoX-2b | 1.7B | F16 | 3.8 GB |
| text-to-video-ms-1.7b | 1.4B | F32 | 6.3 GB |
| LTX-Video-0.9.5 | 1.9B | BF16 | 4.3 GB |
| stable-virtual-camera | 1.3B | F32 | 5.7 GB |
Speech models
The 40 most downloaded of 124.
Embedding models
10 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| Qwen3-Embedding-0.6B | 596M | BF16 | 1.3 GB |
| multilingual-e5-large-instruct | 560M | F16 | 1.3 GB |
| Qwen3-VL-Embedding-2B | 2.1B | BF16 | 4.8 GB |
| gte-Qwen2-1.5B-instruct | 1.8B | F32 | 7.9 GB |
| embeddinggemma-2 | 744M | BF16 | 1.7 GB |
| voyage-4-nano | 346M | BF16 | 0.8 GB |
| jina-code-embeddings-0.5b | 494M | BF16 | 1.1 GB |
| stella_en_1.5B_v5 | 1.5B | F32 | 6.9 GB |
| jina-code-embeddings-1.5b | 1.5B | BF16 | 3.5 GB |
| WeMM-Embedding-2B | 2.7B | BF16 | 6.1 GB |
Other models
The 40 most downloaded of 45.
Fits at FP8
Models that fit only after quantizing the weights to 8 bits (1 byte per parameter). Needs an FP8 checkpoint or an engine that quantizes on load, and a GPU with FP8 support.
Text models
The 40 most downloaded of 310.
Vision-language models
The 40 most downloaded of 93.
Image generation models
The 40 most downloaded of 71.
Video generation models
17 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| LTX-Video | 1.9B | F32 | 2.1 GB |
| Wan2.1-T2V-1.3B-Diffusers | 1.4B | F32 | 1.6 GB |
| stable-video-diffusion-img2vid-xt | 1.5B | F32 | 1.7 GB |
| Wan2.2-TI2V-5B-Diffusers | 5.0B | F32 | 5.6 GB |
| FastWan2.2-TI2V-5B-FullAttn-Diffusers | 5.0B | BF16 | 5.6 GB |
| TurboWan2.1-T2V-1.3B-Diffusers | 1.4B | BF16 | 1.6 GB |
| stable-video-diffusion-img2vid | 1.5B | F32 | 1.7 GB |
| i2vgen-xl | 1.4B | F32 | 1.6 GB |
| Wan2.1-T2V-1.3B | 1.4B | F32 | 1.6 GB |
| CogVideoX-2b | 1.7B | F16 | 1.9 GB |
| text-to-video-ms-1.7b | 1.4B | F32 | 1.6 GB |
| LTX-Video-0.9.5 | 1.9B | BF16 | 2.1 GB |
| CogVideoX-5b | 5.6B | BF16 | 6.2 GB |
| LongLive-2.0-5B-Diffusers | 5.0B | BF16 | 5.6 GB |
| CogVideoX-5b-I2V | 5.6B | BF16 | 6.3 GB |
| Wan2.1-VACE-1.3B-diffusers | 2.2B | F32 | 2.4 GB |
| stable-virtual-camera | 1.3B | F32 | 1.4 GB |
Speech models
The 40 most downloaded of 139.
Embedding models
11 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| Qwen3-Embedding-0.6B | 596M | BF16 | 0.7 GB |
| Qwen3-Embedding-4B | 4.0B | BF16 | 4.5 GB |
| multilingual-e5-large-instruct | 560M | F16 | 0.6 GB |
| Qwen3-VL-Embedding-2B | 2.1B | BF16 | 2.4 GB |
| gte-Qwen2-1.5B-instruct | 1.8B | F32 | 2.0 GB |
| embeddinggemma-2 | 744M | BF16 | 0.8 GB |
| voyage-4-nano | 346M | BF16 | 0.4 GB |
| jina-code-embeddings-0.5b | 494M | BF16 | 0.6 GB |
| stella_en_1.5B_v5 | 1.5B | F32 | 1.7 GB |
| jina-code-embeddings-1.5b | 1.5B | BF16 | 1.7 GB |
| WeMM-Embedding-2B | 2.7B | BF16 | 3.0 GB |
Other models
The 40 most downloaded of 63.
Fits at INT4
Models that fit only after quantizing to 4 bits (0.5 byte per parameter). Needs a quantized checkpoint actually published for the model; check its Hugging Face page before relying on a row.
Text models
The 40 most downloaded of 548.
Vision-language models
The 40 most downloaded of 146.
Image generation models
The 40 most downloaded of 83.
Video generation models
23 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| LTX-Video | 1.9B | F32 | 1.1 GB |
| Wan2.1-T2V-1.3B-Diffusers | 1.4B | F32 | 0.8 GB |
| stable-video-diffusion-img2vid-xt | 1.5B | F32 | 0.9 GB |
| Wan2.2-TI2V-5B-Diffusers | 5.0B | F32 | 2.8 GB |
| Wan2.2-T2V-A14B-Diffusers | 14.3B | F32 | 8.0 GB |
| Wan2.2-I2V-A14B-Diffusers | 14.3B | F32 | 8.0 GB |
| FastWan2.2-TI2V-5B-FullAttn-Diffusers | 5.0B | BF16 | 2.8 GB |
| Wan2.1-T2V-14B-Diffusers | 14.3B | F32 | 8.0 GB |
| TurboWan2.1-T2V-1.3B-Diffusers | 1.4B | BF16 | 0.8 GB |
| stable-video-diffusion-img2vid | 1.5B | F32 | 0.9 GB |
| Wan2.2-I2V-A14B-Lightning-Diffusers | 14.3B | BF16 | 8.0 GB |
| i2vgen-xl | 1.4B | F32 | 0.8 GB |
| Wan2.1-T2V-14B | 14.3B | F32 | 8.0 GB |
| Wan2.1-T2V-1.3B | 1.4B | F32 | 0.8 GB |
| CogVideoX-2b | 1.7B | F16 | 0.9 GB |
| text-to-video-ms-1.7b | 1.4B | F32 | 0.8 GB |
| LTX-Video-0.9.5 | 1.9B | BF16 | 1.1 GB |
| CogVideoX-5b | 5.6B | BF16 | 3.1 GB |
| LongLive-2.0-5B-Diffusers | 5.0B | BF16 | 2.8 GB |
| CogVideoX-5b-I2V | 5.6B | BF16 | 3.1 GB |
| Wan2.1-VACE-1.3B-diffusers | 2.2B | F32 | 1.2 GB |
| stable-virtual-camera | 1.3B | F32 | 0.7 GB |
| mochi-1-preview | 10.0B | F32 | 5.6 GB |
Speech models
The 40 most downloaded of 143.
Embedding models
17 models.
| Model | Parameters | Published as | VRAM needed |
|---|---|---|---|
| Qwen3-Embedding-0.6B | 596M | BF16 | 0.3 GB |
| Qwen3-Embedding-4B | 4.0B | BF16 | 2.2 GB |
| Qwen3-Embedding-8B | 7.6B | BF16 | 4.2 GB |
| multilingual-e5-large-instruct | 560M | F16 | 0.3 GB |
| Qwen3-VL-Embedding-2B | 2.1B | BF16 | 1.2 GB |
| Qwen3-VL-Embedding-8B | 8.1B | BF16 | 4.6 GB |
| gte-Qwen2-1.5B-instruct | 1.8B | F32 | 1.0 GB |
| embeddinggemma-2 | 744M | BF16 | 0.4 GB |
| Qwen3-VL-Embedding-8B-FP8 | 8.8B | F8_E4M3 | 4.9 GB |
| voyage-4-nano | 346M | BF16 | 0.2 GB |
| gte-Qwen2-7B-instruct | 7.6B | F32 | 4.3 GB |
| jina-code-embeddings-0.5b | 494M | BF16 | 0.3 GB |
| stella_en_1.5B_v5 | 1.5B | F32 | 0.9 GB |
| jina-code-embeddings-1.5b | 1.5B | BF16 | 0.9 GB |
| WeMM-Embedding-2B | 2.7B | BF16 | 1.5 GB |
| WeMM-Embedding-9B | 9.4B | BF16 | 5.3 GB |
| pplx-embed-v2-context-9b-preview | 8.4B | F32 | 4.7 GB |
Other models
The 40 most downloaded of 88.
How these numbers are computed
Required VRAM is the weight size at each precision times a flat 1.2 overhead, the same figure every model page shows. It does not include a long context: the KV cache grows with every token, so a model near the top of a tier can need the next one at long context. Read how much VRAM you need for LLMs for the method, the VRAM and quantization glossary entries for the terms, and the VRAM calculator to size a model that is not listed.
Looking for a model by job rather than by memory? Start with coding, reasoning, chat and assistants, vision-language or see the full models directory.