What can the RTX 4070 Super run?
| Text models: 246 | Visual models: 12 | Strongest first |
- Memory
- 12 GB
- Bandwidth
- 504 GB/s
- Maker
- NVIDIA
Open the calculator on this machineEvery model for 12 GB of VRAM
Text models that run on it
| Model | Best fit | Context | Speed | Strongest at |
|---|---|---|---|---|
| Qwen3.5 9B | 8-bit | 24K | 37 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 12B | 6-bit | 112K | 34 tok/s | Chat, Coding, Docs |
| Qwen3.5 4B | 8-bit | 176K | 66 tok/s | Agents, Docs |
| Ministral 3 14B | 4-bit | 16K | 42 tok/s | Coding, Agents, Docs |
| Gemma 4 E4B | 8-bit | 128K | 59 tok/s | Chat, Coding, Docs |
| gpt-oss 20B | 3-bit | 40K | 156 tok/s | Agents |
| Ministral 3 8B | 6-bit | 28K | 49 tok/s | Docs |
| Gemma 4 E2B | 8-bit | 128K | 106 tok/s | Chat |
| Nemotron 3 Nano 4B | 8-bit | 256K | 82 tok/s | Docs |
| Llama 3.1 8B | 8-bit | 20K | 42 tok/s | Chat, Coding, Agents, Docs |
| Qwen3 0.6B | 8-bit | 40K | 322 tok/s | Not rated |
| Qwen3 14B | 3-bit | 24K | 49 tok/s | Not rated |
| Qwen3 VL 8B Instruct | 6-bit | 24K | 50 tok/s | Text and images |
| Qwen3 8B | 8-bit | 16K | 42 tok/s | Not rated |
| Qwen2.5 7B Instruct | 8-bit | 32K | 45 tok/s | Not rated |
| Qwen2.5 VL 7B Instruct | 8-bit | 40K | 41 tok/s | Text and images |
| Unlimited OCR | 8-bit | 32K | 241 tok/s | Reading documents |
| Qwen2.5 1.5B Instruct | 8-bit | 32K | 182 tok/s | Not rated |
| Qwen3.5 2B | 8-bit | 256K | 131 tok/s | Text and images |
| Qwen3 VL 4B Instruct | 8-bit | 40K | 69 tok/s | Text and images |
| Qwen3 1.7B | 8-bit | 40K | 144 tok/s | Not rated |
| Qwen3 4B | 8-bit | 40K | 76 tok/s | Not rated |
| Qwen3 4B Instruct 2507 | 8-bit | 44K | 76 tok/s | Not rated |
| Qwen2.5 3B Instruct | 8-bit | 32K | 96 tok/s | Not rated |
| DeepSeek OCR | 8-bit | 8K | 241 tok/s | Reading documents |
| Qwen2.5 Coder 14B Instruct | 3-bit | 20K | 48 tok/s | Coding |
| Qwen2.5 VL 3B Instruct | 8-bit | 125K | 81 tok/s | Text and images |
| Qwen2.5 Coder 7B Instruct | 8-bit | 32K | 45 tok/s | Coding |
| Mistral 7B Instruct V0.2 | 8-bit | 28K | 45 tok/s | Not rated |
| Chandra OCR 2 | 8-bit | 160K | 59 tok/s | Reading documents |
| Qwen3 VL 2B Instruct | 8-bit | 72K | 138 tok/s | Text and images |
| Qwen2 VL 2B Instruct | 8-bit | 32K | 133 tok/s | Text and images |
| MiniCPM5 2B | 8-bit | 128K | 124 tok/s | Not rated |
| Gemma 3 4B IT | 8-bit | 128K | 84 tok/s | Text and images |
| Llama 3.2 3B Instruct | 8-bit | 64K | 95 tok/s | Not rated |
| LightOnOCR 2 1B | 8-bit | 16K | 258 tok/s | Text and images |
| Qwen2.5 14B Instruct | 3-bit | 20K | 48 tok/s | Not rated |
| Qwen3 1.7B Base | 8-bit | 32K | 166 tok/s | Not rated |
| HunyuanOCR | 8-bit | 96K | 244 tok/s | Text and images |
| Qwen3 0.6B Base | 8-bit | 32K | 379 tok/s | Not rated |
| Rax 4.5 | 8-bit | 256K | 131 tok/s | Text and images |
| Qwen2 VL 7B Instruct | 8-bit | 32K | 41 tok/s | Text and images |
| Ornith 1.5 9B | 8-bit | 20K | 37 tok/s | Not rated |
| Unlimited OCR 2 | 8-bit | 120K | 239 tok/s | Reading documents |
| Granite 4.1 3B | 8-bit | 88K | 87 tok/s | Not rated |
| DeepSeek R1 Distill Qwen 1.5B | 8-bit | 128K | 182 tok/s | Not rated |
| Mistral 7B V0.1 | 8-bit | 32K | 45 tok/s | Not rated |
| Qwen2.5 Coder 3B Instruct | 8-bit | 32K | 96 tok/s | Coding |
| Meta Llama 3 8B Instruct Abliterated V3 | 8-bit | 8K | 42 tok/s | Not rated |
| EXAONE 3.5 7.8B Instruct | 8-bit | 20K | 43 tok/s | Not rated |
| Saiga Llama3 8B | 8-bit | 8K | 42 tok/s | Not rated |
| Qwen2.5 Coder 1.5B Instruct | 8-bit | 32K | 182 tok/s | Coding |
| DeepSeek R1 0528 Qwen3 8B | 8-bit | 16K | 42 tok/s | Not rated |
| Gemma 3 12B IT | 5-bit | 28K | 42 tok/s | Text and images |
| Phi 3.5 Mini Instruct | 8-bit | 16K | 82 tok/s | Not rated |
| Qwen3 4B Thinking 2507 | 8-bit | 44K | 76 tok/s | Not rated |
| Phi 4 Mini Instruct | 8-bit | 52K | 80 tok/s | Not rated |
| Dots.ocr | 8-bit | 128K | 108 tok/s | Reading documents |
| DeepSeek R1 Distill Qwen 7B | 8-bit | 52K | 45 tok/s | Not rated |
| Phi 3 Mini 4K Instruct | 8-bit | 4K | 82 tok/s | Not rated |
Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 246 are shown.
Image and video models
| Model | Best fit | Per output | Strongest at |
|---|---|---|---|
| Stable Diffusion XL 1.0 | 16-bit | 9.6 s | Images |
| Z-Image Turbo | 8-bit | 13 s | Images |
| FLUX.1 [dev] | 6-bit | 30 s | Images |
| Qwen-Image 2.1 | 8-bit | 3.5 min | Images |
| FLUX.3 Klein | 16-bit | 19 s | Images |
| FLUX.2 [klein] 9B | 8-bit | 6.8 s | Images |
| Krea 2 Turbo | 5-bit | 30 s | Images |
| Stable Diffusion 3.5 Large | 8-bit | 1.8 min | Images |
| MOSS-Image | 16-bit | 2.3 min | Images |
| Wan 2.2 TI2V-5B | 16-bit | 20 min | Video |
| MOSS Image Base | 16-bit | 2.3 min | Images |
| HunyuanVideo 1.5 | 6-bit | 1.7 h | Video |
Only just too big
- Qwen3.8 27B
- Muse Glimmer 30B
- Gemma 4 26B A4B
- Devstral Small 2 24B
- LFM2.5 24B A2B
- Mistral Small 3.2 24B
- MiniMax H3
- LTX-2.5
- Qwen3.6 27B
- Qwen3.5 27B
- Diffusiongemma 26B A4B IT
- Gemma 3 27B IT
- Qwen-Image
- GPT Neox 20B
- Wan2.2 I2V A14B
- CodeLlama 7B HF
- Wan2.1 I2V 14B
- Wan 2.1 T2V-14B
- Qwen1.5 7B
- Deepseek Coder 6.7B Base
- Qwen 7B Chat
- Qwen Image 2512
- Molmo2 O 7B
- Internlm2 Chat 20B
- Qwen Image Bench
- Internlm2 Chat 20B ExPO
- Medgemma 27B Text IT
- Qwen1.5 14B Chat
- CodeLlama 13B Instruct HF
- Wan 2.2 A14B
These need up to a quarter more memory than the RTX 4070 Super has. Spilling into system RAM or a lower precision can get them running, slowly.
Questions
What is the best LLM for the RTX 4070 Super?
Of the open models we track, the strongest that run well on the RTX 4070 Super are Qwen3.5 9B, Gemma 4 12B, Qwen3.5 4B. 246 text models fit it in all.
Calculated from published specs, not benchmarked. All machines