Best LLMs for 8 GB of VRAM
| Text models: 234 | Visual models: 6 | Sized for RTX 4060 | Strongest first |
Text models for 8 GB
| Model | Best fit | Context | Speed | Strongest at |
|---|---|---|---|---|
| Qwen3.5 9B | 4-bit | 32K | 35 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 12B | 3-bit | 96K | 31 tok/s | Chat, Coding, Docs |
| Qwen3.5 4B | 8-bit | 60K | 37 tok/s | Agents, Docs |
| Gemma 4 E4B | 6-bit | 28K | 42 tok/s | Chat, Coding, Docs |
| Ministral 3 8B | 3-bit | 20K | 44 tok/s | Docs |
| Gemma 4 E2B | 8-bit | 128K | 60 tok/s | Chat |
| Nemotron 3 Nano 4B | 8-bit | 176K | 46 tok/s | Docs |
| Llama 3.1 8B | 4-bit | 16K | 40 tok/s | Chat, Coding, Agents, Docs |
| Qwen3 0.6B | 8-bit | 40K | 199 tok/s | Not rated |
| Qwen3 VL 8B Instruct | 3-bit | 16K | 45 tok/s | Text and images |
| Qwen3 8B | 3-bit | 20K | 49 tok/s | Not rated |
| Qwen2.5 7B Instruct | 6-bit | 16K | 32 tok/s | Not rated |
| Qwen2.5 VL 7B Instruct | 5-bit | 20K | 34 tok/s | Text and images |
| Unlimited OCR | 8-bit | 32K | 141 tok/s | Reading documents |
| Qwen2.5 1.5B Instruct | 8-bit | 32K | 106 tok/s | Not rated |
| Qwen3.5 2B | 8-bit | 256K | 74 tok/s | Text and images |
| Qwen3 VL 4B Instruct | 8-bit | 16K | 39 tok/s | Text and images |
| Qwen3 1.7B | 8-bit | 40K | 82 tok/s | Not rated |
| Qwen3.5 0.8B | 8-bit | 256K | 176 tok/s | Text and images |
| Qwen3 4B | 8-bit | 16K | 43 tok/s | Not rated |
| Qwen3 4B Instruct 2507 | 8-bit | 16K | 43 tok/s | Not rated |
| Qwen2.5 3B Instruct | 8-bit | 32K | 55 tok/s | Not rated |
| DeepSeek OCR | 8-bit | 8K | 141 tok/s | Reading documents |
| Qwen2.5 VL 3B Instruct | 8-bit | 80K | 45 tok/s | Text and images |
| Qwen2.5 Coder 7B Instruct | 6-bit | 16K | 32 tok/s | Coding |
| Mistral 7B Instruct V0.2 | 5-bit | 16K | 36 tok/s | Not rated |
| Chandra OCR 2 | 8-bit | 40K | 32 tok/s | Reading documents |
| Llama 3.2 1B Instruct | 8-bit | 128K | 136 tok/s | Not rated |
| Qwen3 VL 2B Instruct | 8-bit | 40K | 79 tok/s | Text and images |
| Qwen2 VL 2B Instruct | 8-bit | 32K | 76 tok/s | Text and images |
| MiniCPM5 2B | 8-bit | 104K | 72 tok/s | Not rated |
| Gemma 3 4B IT | 8-bit | 112K | 47 tok/s | Text and images |
| Llama 3.2 3B Instruct | 8-bit | 32K | 53 tok/s | Not rated |
| LightOnOCR 2 1B | 8-bit | 16K | 155 tok/s | Text and images |
| Qwen3 1.7B Base | 8-bit | 32K | 96 tok/s | Not rated |
| HunyuanOCR | 8-bit | 60K | 143 tok/s | Text and images |
| Qwen3 0.6B Base | 8-bit | 32K | 240 tok/s | Not rated |
| Rax 4.5 | 8-bit | 256K | 74 tok/s | Text and images |
| Qwen2 VL 7B Instruct | 5-bit | 20K | 34 tok/s | Text and images |
| Ornith 1.5 9B | 4-bit | 32K | 35 tok/s | Not rated |
| Unlimited OCR 2 | 8-bit | 60K | 140 tok/s | Reading documents |
| Granite 4.1 3B | 8-bit | 44K | 50 tok/s | Not rated |
| DeepSeek R1 Distill Qwen 1.5B | 8-bit | 128K | 106 tok/s | Not rated |
| Mistral 7B V0.1 | 6-bit | 32K | 32 tok/s | Not rated |
| Qwen2.5 Coder 3B Instruct | 8-bit | 32K | 55 tok/s | Coding |
| Meta Llama 3 8B Instruct Abliterated V3 | 5-bit | 8K | 34 tok/s | Not rated |
| EXAONE 3.5 7.8B Instruct | 4-bit | 16K | 41 tok/s | Not rated |
| Saiga Llama3 8B | 5-bit | 8K | 34 tok/s | Not rated |
| MiniCPM V 4.7 1B | 8-bit | 256K | 124 tok/s | Text and images |
| Qwen2.5 Coder 1.5B Instruct | 8-bit | 32K | 106 tok/s | Coding |
| DeepSeek R1 0528 Qwen3 8B | 3-bit | 20K | 49 tok/s | Not rated |
| Qwen3 4B Thinking 2507 | 8-bit | 16K | 43 tok/s | Not rated |
| Phi 4 Mini Instruct | 8-bit | 20K | 45 tok/s | Not rated |
| Dots.ocr | 8-bit | 128K | 61 tok/s | Reading documents |
| DeepSeek R1 Distill Qwen 7B | 6-bit | 16K | 32 tok/s | Not rated |
| Phi 3 Mini 4K Instruct | 8-bit | 4K | 46 tok/s | Not rated |
| DeepSeek OCR 2 | 8-bit | 8K | 135 tok/s | Reading documents |
| SmolLM3 3B | 8-bit | 52K | 55 tok/s | Not rated |
| Qwen2 7B Instruct | 6-bit | 16K | 32 tok/s | Not rated |
| Apertus 8B Instruct 2509 | 4-bit | 16K | 40 tok/s | Not rated |
Fits, context and speed on the RTX 4060. A slower card with the same memory holds the same models at lower speed. Ranked by our use-case scores, then by Hugging Face downloads. The top 60 of 234 are shown.
Image and video models
| Model | Best fit | Per output | Strongest at |
|---|---|---|---|
| Stable Diffusion XL 1.0 | 16-bit | 21 s | Images |
| Z-Image Turbo | 8-bit | 30 s | Images |
| FLUX.1 [dev] | 4-bit | 69 s | Images |
| FLUX.3 Klein | 8-bit | 44 s | Images |
| Stable Diffusion 3.5 Large | 5-bit | 4.9 min | Images |
| Wan 2.2 TI2V-5B | 8-bit | 46 min | Video |
Machines with 8 GB
Questions
What is the best LLM for 8 GB of VRAM?
Of the open models we track, the strongest that run well on 8 GB are Qwen3.5 9B, Gemma 4 12B, Qwen3.5 4B. 234 text models fit in all.
Calculated from published specs, not benchmarked. Every machine and tier