The RTX 5080 is the fastest 16GB card for local AI. At 94 tokens per second for 8B models, it outperforms even the RTX 3090 in raw speed while costing less. The best choice for users who want top speed with 14B models.
| GPU | VRAM | Speed | Bandwidth | Price |
|---|---|---|---|---|
| RTX 5070 Ti | 16 GB | 87 tok/s | 896 GB/s | $749 |
| RTX 5080 | 16 GB | 94 tok/s | 960 GB/s | $999 |
| RTX 4080 SUPER | 16 GB | 79 tok/s | 736 GB/s | $1,597 |
| RTX 5090 | 32 GB | 145 tok/s | 1792 GB/s | $2,499 |
Llama / 8B / Q4_K_M / ~6.5 GB
Best for: Chat, Coding·Pop: 94/100
Perf: ~94.0 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run llama3.1:8b-instruct-q4_K_M
Qwen / 9B / Q4_K_M / ~7 GB
Best for: Quality, Coding, Reasoning·Pop: 86/100
Perf: ~85.0 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for quality, coding, reasoning on RTX 5080.
ollama run qwen3.5:9b-instruct-q4_K_M
Qwen / 8B / Q4_K_M / ~6.5 GB
Best for: Chat, Coding·Pop: 88/100
Perf: ~94.0 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run qwen3:8b-q4_K_M
Mistral / 7B / Q4_K_M / ~5.5 GB
Best for: Chat, Coding·Pop: 90/100
Perf: ~105.3 tok/s · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run mistral:7b-instruct-q4_K_M
Qwen / 7B / Q4_K_M / ~5.5 GB
Best for: Coding·Pop: 85/100
Perf: ~105.3 tok/s · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for coding on RTX 5080.
ollama run qwen2.5-coder:7b-q4_K_M
Qwen / 7B / Q4_K_M / ~5.5 GB
Best for: Chat, Coding·Pop: 86/100
Perf: ~105.3 tok/s · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run qwen2.5:7b-instruct-q4_K_M
LFM2 / 8B / Q4_K_M / ~6 GB
Best for: Local agents, tool calling, fast chat·Pop: 75/100
Perf: ~94.0 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for local agents, tool calling, fast chat on RTX 5080.
ollama run liquidai/lfm2:8b-a1b-instruct-q4_K_M
DeepSeek / 7B / Q4_K_M / ~5.5 GB
Best for: Reasoning, Coding·Pop: 77/100
Perf: ~105.3 tok/s · first token ~0.3s
Fits in 16 GB VRAM with room to spare. Best for reasoning, coding on RTX 5080.
ollama run deepseek-r1-distill:qwen-7b-q4_K_M
Llama / 8B / Q5_K_M / ~8 GB
Best for: Chat, Coding·Pop: 82/100
Perf: ~80.8 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run llama3.1:8b-instruct-q5_K_M
Gemma / 9B / Q4_K_M / ~7 GB
Best for: Chat, Coding·Pop: 81/100
Perf: ~85.0 tok/s · first token ~0.4s
Fits in 16 GB VRAM with room to spare. Best for chat, coding on RTX 5080.
ollama run gemma2:9b-instruct-q4_K_M
With 16GB VRAM, the RTX 5080 can run up to 14b parameter models. Top recommendations include Llama 3.1 8B Instruct, Qwen3.5 9B Instruct, Qwen3 8B.
The RTX 5080 achieves 94 tokens per second with Qwen3 8B at Q4 quantization. Smaller models run faster, larger models slower.
16GB VRAM is good for local AI. You can comfortably run up to 14b parameter models with room for KV cache. 10 of our top 10 recommended models run at full speed.
Install Ollama from ollama.com, then run models directly. For example: ollama run llama3.1:8b-instruct-q4_K_M. Ollama automatically detects your NVIDIA GPU and uses CUDA acceleration.
Use our interactive wizard to compare models across Apple Silicon and NVIDIA GPUs.
Open ModelFit Wizard →